{"meta":{"count":174,"source":"src/data"},"items":[{"id":"frontier-chat","slug":"frontier-chat","name":"Frontier general chat & reasoning","shortName":"Frontier general chat & reasoning","problemStatement":"Need the strongest general-purpose LLM API for product chat, planning, and hard reasoning.","whenNotTopPick":"Need best multimodal media APIs in one vendor (OpenAI/Google), or procurement forces Bedrock/Azure/Vertex.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"anthropic","depotPickRationale":"Weighted capability + reliability + privacy. Anthropic Sonnet/Opus lead agentic reasoning; OpenAI closest peer; Gemini for value/multimodal.","runnersUpIds":["openai","google-ai","xai","minimax","dashscope","zai"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"Weighted capability + reliability + privacy. Anthropic Sonnet/Opus lead agentic reasoning; OpenAI closest peer; Gemini for value/multimodal.","tags":["frontier"]},{"id":"coding-agent","slug":"coding-agent","name":"Coding agent / code generation","shortName":"Coding agent / code generation","problemStatement":"Agent must edit repos, call tools, and produce correct code with high reliability.","whenNotTopPick":"Compliance blocks Anthropic; or you need cheapest high-volume codegen (DeepSeek/open hosts) with human review.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"anthropic","depotPickRationale":"Capability and reliability dominate. PricePerf secondary for interactive coding; primary for bulk codegen.","runnersUpIds":["openai","deepseek","mistral","zai","minimax","dashscope"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability and reliability dominate. PricePerf secondary for interactive coding; primary for bulk codegen.","tags":["coding"]},{"id":"long-context-agents","slug":"long-context-agents","name":"Long-context agent workflows","shortName":"Long-context agent workflows","problemStatement":"Need 200K–1M context for large docs, codebases, or multi-hour agent state.","whenNotTopPick":"Quality on tool use matters more than raw context — then Anthropic; enterprise GCP already → Vertex; want Kimi K3 1M ctx at $3/$15 — moonshot.","categoryIds":["cat-inference-apis","cat-media-gen"],"depotPickProviderId":"google-ai","depotPickRationale":"Capability includes context window + needle quality. Gemini often wins value at 1M; Claude strong at high quality long context.","runnersUpIds":["anthropic","openai","moonshot","ai21"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability includes context window + needle quality. Gemini often wins value at 1M; Claude strong at high quality long context.","tags":["long"]},{"id":"agent-tools","slug":"agent-tools","name":"Tool-calling production agents","shortName":"Tool-calling production agents","problemStatement":"Reliable structured tool use, parallel tools, and low schema-slip in production agents.","whenNotTopPick":"You need OpenAI Realtime/multimodal tools specifically, or a multi-provider router during migration (OpenRouter).","categoryIds":["cat-inference-apis"],"depotPickProviderId":"anthropic","depotPickRationale":"Capability + reliability + dx. Prefer first-party APIs over routers for SLA-critical tool loops.","runnersUpIds":["openai","google-ai","openrouter"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability + reliability + dx. Prefer first-party APIs over routers for SLA-critical tool loops.","tags":["agent"]},{"id":"cheap-volume-llm","slug":"cheap-volume-llm","name":"High-volume cheap LLM inference","shortName":"High-volume cheap LLM inference","problemStatement":"Millions of tokens/day for classification, extraction, or drafts where frontier quality is optional.","whenNotTopPick":"Need absolute lowest latency on a supported model (Groq/Cerebras) or compliance requires Bedrock/Azure.","categoryIds":["cat-inference-apis","cat-train-finetune"],"depotPickProviderId":"fireworks","depotPickRationale":"pricePerf + reliability + dx. Open-model hosts win; Gemini Flash also excellent closed-model value.","runnersUpIds":["together","groq","deepinfra","deepseek","google-ai","dashscope","zai","chutes"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf + reliability + dx. Open-model hosts win; Gemini Flash also excellent closed-model value.","tags":["cheap"]},{"id":"open-model-inference","slug":"open-model-inference","name":"Hosted open-weight model inference","shortName":"Hosted open-weight model inference","problemStatement":"Run Llama/Qwen/DeepSeek/etc. without owning GPUs, with production DX.","whenNotTopPick":"Need Hub-native custom model endpoints (HF) or ultra-low latency on narrow SKUs (Groq).","categoryIds":["cat-inference-apis","cat-train-finetune"],"depotPickProviderId":"fireworks","depotPickRationale":"Catalog freshness, tokens/sec, OpenAI-compat DX, and pricePerf.","runnersUpIds":["together","groq","deepinfra","huggingface-inference","chutes","friendli"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Catalog freshness, tokens/sec, OpenAI-compat DX, and pricePerf.","tags":["open"]},{"id":"ultra-low-latency","slug":"ultra-low-latency","name":"Ultra-low-latency token streaming","shortName":"Ultra-low-latency token streaming","problemStatement":"Voice/realtime UX needs extremely fast tokens/sec on supported models.","whenNotTopPick":"Required model isn't on Groq; or you need frontier Claude/GPT quality more than speed.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"groq","depotPickRationale":"reliability of latency + pricePerf. Model coverage is the hard constraint.","runnersUpIds":["cerebras","sambanova","fireworks","friendli"],"weights":{"capabilityFit":0.2,"pricePerformance":0.15,"latencyReliability":0.3,"developerDx":0.15,"privacyPosture":0.08,"lockInRisk":0.05,"maturity":0.07},"methodologyNotes":"reliability of latency + pricePerf. Model coverage is the hard constraint.","tags":["ultra"]},{"id":"multi-model-router","slug":"multi-model-router","name":"Multi-provider model routing","shortName":"Multi-provider model routing","problemStatement":"One integration to A/B models and fail over across vendors.","whenNotTopPick":"Need a single-vendor enterprise SLA — pin Anthropic/OpenAI or use Bedrock/Vertex as the governed router.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"openrouter","depotPickRationale":"dx + capability breadth; reliability discounted because upstream variance.","runnersUpIds":["amazon-bedrock","vertex-ai","together","portkey"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + capability breadth; reliability discounted because upstream variance.","tags":["multi"]},{"id":"enterprise-llm","slug":"enterprise-llm","name":"Enterprise-governed LLM access","shortName":"Enterprise-governed LLM access","problemStatement":"Procurement, VPC, IAM, BAA, and regional residency dictate cloud wrappers.","whenNotTopPick":"Team can sign Anthropic/OpenAI direct with ZDR — often better DX. Azure-only shops → Azure OpenAI.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"amazon-bedrock","depotPickRationale":"privacy + maturity + reliability over raw dx/pricePerf.","runnersUpIds":["azure-openai","vertex-ai","anthropic"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"privacy + maturity + reliability over raw dx/pricePerf.","tags":["enterprise"]},{"id":"eu-sovereign-llm","slug":"eu-sovereign-llm","name":"EU-friendly / sovereign LLM","shortName":"EU-friendly / sovereign LLM","problemStatement":"Prefer EU vendor posture and data residency options for LLM APIs.","whenNotTopPick":"Need absolute frontier quality regardless of vendor HQ — Anthropic/OpenAI with EU regions/ZDR; need German Schwarz-cloud sovereignty — stackit-ai; Paris open-model serverless — scaleway-generative; Swiss public open-model API — infomaniak; need to run your own weights on PL/DE GPUs — cloudferro; need on-prem DACH PhariaOS and will wait 4–12 weeks — aleph-alpha.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"mistral","depotPickRationale":"privacy + regions + maturity; capability tempered by sovereignty constraint.","runnersUpIds":["scaleway-generative","ionos-ai","t-systems-aifs","ovhcloud-ai","stackit-ai","infomaniak","aleph-alpha","cloudferro"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"privacy + regions + maturity; capability tempered by sovereignty constraint.","tags":["eu"]},{"id":"embeddings","slug":"embeddings","name":"Embeddings for RAG","shortName":"Embeddings for RAG","problemStatement":"High-quality embeddings (+ optional rerank) for retrieval pipelines.","whenNotTopPick":"Already standardized on OpenAI embeddings operationally; retrieval-quality bake-off winner is Voyage; need local/open embed models on HF/Fireworks; want URL-to-markdown + embeddings together — Jina.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"cohere","depotPickRationale":"Capability for retrieval quality + pricePerf. Cohere remains general default (enterprise + multilingual). Voyage when nDCG/lite $0.02 SKU wins the bake-off; Jina when you also want Reader/rerank in one EU API.","runnersUpIds":["voyage","openai","google-ai","jina","huggingface-inference"],"weights":{"capabilityFit":0.26,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.22,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"Capability for retrieval quality + pricePerf. Cohere remains general default (enterprise + multilingual). Voyage when nDCG/lite $0.02 SKU wins the bake-off; Jina when you also want Reader/rerank in one EU API.","tags":["embeddings"]},{"id":"rag-enterprise","slug":"rag-enterprise","name":"Enterprise RAG stack","shortName":"Enterprise RAG stack","problemStatement":"Managed retrieval quality with rerank, security, and support.","whenNotTopPick":"Simple app RAG on Postgres — use supabase-pgvector instead of a dedicated RAG vendor.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"cohere","depotPickRationale":"privacy + capability + maturity.","runnersUpIds":["pinecone","weaviate","amazon-bedrock"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"privacy + capability + maturity.","tags":["rag"]},{"id":"multimodal","slug":"multimodal","name":"Multimodal understanding (vision+)","shortName":"Multimodal understanding (vision+)","problemStatement":"Agents must reason over images/video/audio inputs at scale.","whenNotTopPick":"Text-only coding agents — Anthropic; or OpenAI Realtime voice stacks.","categoryIds":["cat-inference-apis","cat-media-gen"],"depotPickProviderId":"google-ai","depotPickRationale":"capability multimodal + pricePerf. Gemini often wins value; OpenAI strong ecosystem.","runnersUpIds":["openai","anthropic","minimax","dashscope","zai"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability multimodal + pricePerf. Gemini often wins value; OpenAI strong ecosystem.","tags":["multimodal"]},{"id":"web-grounded-qa","slug":"web-grounded-qa","name":"Web-grounded Q&A with citations","shortName":"Web-grounded Q&A with citations","problemStatement":"Answers must reflect live web with citations without DIY search stack.","whenNotTopPick":"Need full control of retrieval/ranking — compose Exa/Tavily + your LLM.","categoryIds":["cat-inference-apis","cat-search-scrape"],"depotPickProviderId":"perplexity","depotPickRationale":"capability groundedness + dx; lock-in medium.","runnersUpIds":["exa","tavily","openai","you-com"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability groundedness + dx; lock-in medium.","tags":["web"]},{"id":"serverless-gpu","slug":"serverless-gpu","name":"Serverless GPU inference/jobs","shortName":"Serverless GPU inference/jobs","problemStatement":"Run GPU work with scale-to-zero and minimal ops.","whenNotTopPick":"Need cheapest raw GPU-seconds on familiar pods (RunPod) or one-click public models (Replicate).","categoryIds":["cat-gpu-cloud","cat-cpu-compute"],"depotPickProviderId":"modal","depotPickRationale":"dx + pricePerf + reliability. Modal wins Python DX; RunPod wins flexible $/GPU.","runnersUpIds":["runpod","replicate","beam","baseten","chutes","friendli","koyeb"],"weights":{"capabilityFit":0.22,"pricePerformance":0.25,"latencyReliability":0.15,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"dx + pricePerf + reliability. Modal wins Python DX; RunPod wins flexible $/GPU.","tags":["serverless"]},{"id":"cheap-gpu","slug":"cheap-gpu","name":"Cheapest on-demand GPUs","shortName":"Cheapest on-demand GPUs","problemStatement":"Minimize $/GPU-hr for fault-tolerant training/batch.","whenNotTopPick":"Need SLA/security — RunPod Secure Cloud, Lambda, or CoreWeave. Never Vast for sensitive PII without review.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"vast-ai","depotPickRationale":"pricePerf primary; reliability heavily discounted for marketplaces.","runnersUpIds":["runpod","oracle-gpu","fluidstack","thunder-compute","verda"],"weights":{"capabilityFit":0.22,"pricePerformance":0.25,"latencyReliability":0.15,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"pricePerf primary; reliability heavily discounted for marketplaces.","tags":["cheap"]},{"id":"predictable-gpu","slug":"predictable-gpu","name":"Predictable dedicated GPU capacity","shortName":"Predictable dedicated GPU capacity","problemStatement":"Labs need known capacity and clusters without marketplace chaos.","whenNotTopPick":"Need max enterprise K8s (CoreWeave) or burst serverless (Modal/RunPod).","categoryIds":["cat-gpu-cloud","cat-train-finetune"],"depotPickProviderId":"lambda","depotPickRationale":"reliability + maturity + capability (interconnect).","runnersUpIds":["coreweave","crusoe","runpod","verda","massed-compute","hyperstack"],"weights":{"capabilityFit":0.22,"pricePerformance":0.25,"latencyReliability":0.15,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"reliability + maturity + capability (interconnect).","tags":["predictable"]},{"id":"multi-node-training","slug":"multi-node-training","name":"Multi-node GPU training","shortName":"Multi-node GPU training","problemStatement":"Train/fine-tune large models across many GPUs with fast interconnect.","whenNotTopPick":"Single-node LoRA — RunPod/Modal. Already all-in on AWS — P5/SageMaker.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"coreweave","depotPickRationale":"capability interconnect + reliability + maturity; pricePerf secondary.","runnersUpIds":["lambda","aws-gpu","crusoe"],"weights":{"capabilityFit":0.26,"pricePerformance":0.22,"latencyReliability":0.1,"developerDx":0.18,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"capability interconnect + reliability + maturity; pricePerf secondary.","tags":["multi"]},{"id":"enterprise-gpu","slug":"enterprise-gpu","name":"Enterprise GPU cloud","shortName":"Enterprise GPU cloud","problemStatement":"SOC2, dedicated capacity, K8s, procurement-friendly GPU estate.","whenNotTopPick":"Indie team without enterprise needs — RunPod/Lambda.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"coreweave","depotPickRationale":"privacy + maturity + reliability.","runnersUpIds":["aws-gpu","gcp-gpu","azure-gpu","crusoe"],"weights":{"capabilityFit":0.22,"pricePerformance":0.25,"latencyReliability":0.15,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"privacy + maturity + reliability.","tags":["enterprise"]},{"id":"production-inference-fleet","slug":"production-inference-fleet","name":"Self-hosted production inference fleet","shortName":"Self-hosted production inference fleet","problemStatement":"Always-on model serving on rented GPUs with control over stacks.","whenNotTopPick":"Prefer tokens not GPUs — Fireworks/Together. Huge enterprise K8s — CoreWeave/AWS.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"runpod","depotPickRationale":"reliability + dx + pricePerf balance.","runnersUpIds":["coreweave","modal","baseten","aws-gpu","friendli"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"reliability + dx + pricePerf balance.","tags":["production"]},{"id":"batch-gpu","slug":"batch-gpu","name":"Batch / interruptible GPU jobs","shortName":"Batch / interruptible GPU jobs","problemStatement":"Offline evals, embeddings, or training steps that tolerate preemption.","whenNotTopPick":"Strict deadlines without restart logic — dedicated on-demand.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"runpod","depotPickRationale":"pricePerf + dx for spot/serverless batch.","runnersUpIds":["vast-ai","modal","aws-gpu"],"weights":{"capabilityFit":0.22,"pricePerformance":0.25,"latencyReliability":0.15,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"pricePerf + dx for spot/serverless batch.","tags":["batch"]},{"id":"finetune","slug":"finetune","name":"Fine-tune open models","shortName":"Fine-tune open models","problemStatement":"SFT/DPO/LoRA without building training infra from scratch.","whenNotTopPick":"Many LoRA adapters on shared GPUs — Predibase/LoRAX; full custom training clusters — Lambda/CoreWeave; Hub-centric — HF; Lightning/PyTorch Studio DX — lightning-ai. Do not start new work on OpenPipe hosted (migrated/sunset).","categoryIds":["cat-inference-apis","cat-train-finetune"],"depotPickProviderId":"together","depotPickRationale":"capability fine-tune features + pricePerf + dx.","runnersUpIds":["unsloth","fireworks","predibase","huggingface-inference","modal","lightning-ai"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability fine-tune features + pricePerf + dx.","tags":["finetune"]},{"id":"app-hosting","slug":"app-hosting","name":"Host agent web app / API","shortName":"Host agent web app / API","problemStatement":"Deploy the product API and UI with minimal ops.","whenNotTopPick":"Next.js UI-heavy — Vercel; global edge Machines — Fly; enterprise AWS — EC2/ECS.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"railway","depotPickRationale":"dx + reliability + pricePerf for early/mid stage.","runnersUpIds":["render","fly-io","vercel","digitalocean"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + reliability + pricePerf for early/mid stage.","tags":["app"]},{"id":"agent-backend","slug":"agent-backend","name":"Long-running agent worker backend","shortName":"Long-running agent worker backend","problemStatement":"Workers that run longer than serverless HTTP timeouts.","whenNotTopPick":"Already on K8s/EC2 with good orchestration — stay. Pure edge short tools — CF Workers.","categoryIds":["cat-gpu-cloud","cat-cpu-compute"],"depotPickProviderId":"modal","depotPickRationale":"capability for long jobs + dx + reliability.","runnersUpIds":["convex","fly-io","railway","render","aws-ec2"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability for long jobs + dx + reliability.","tags":["agent"]},{"id":"edge-deploy","slug":"edge-deploy","name":"Global edge deploy for agent gateways","shortName":"Global edge deploy for agent gateways","problemStatement":"Low-latency thin agent gateways and tool endpoints worldwide.","whenNotTopPick":"Heavy CPU/GPU in request path — Modal/Fly Machines.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"cloudflare-workers","depotPickRationale":"reliability + pricePerf + global regions.","runnersUpIds":["fly-io","vercel","deno-deploy"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"reliability + pricePerf + global regions.","tags":["edge"]},{"id":"cheap-vms","slug":"cheap-vms","name":"Cheap always-on VMs","shortName":"Cheap always-on VMs","problemStatement":"Minimize cost for 24/7 CPU boxes (bots, queues, reverse proxies).","whenNotTopPick":"Need many global regions or enterprise IAM — hyperscalers.","categoryIds":["cat-cpu-compute","cat-object-storage"],"depotPickProviderId":"hetzner","depotPickRationale":"pricePerf + maturity; dx less critical.","runnersUpIds":["digitalocean","oracle-gpu","aws-ec2"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf + maturity; dx less critical.","tags":["cheap"]},{"id":"enterprise-compute","slug":"enterprise-compute","name":"Enterprise general compute","shortName":"Enterprise general compute","problemStatement":"VMs inside an existing hyperscaler estate with compliance.","whenNotTopPick":"Greenfield indie — Hetzner/Fly/Railway.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"aws-ec2","depotPickRationale":"maturity + privacy + reliability.","runnersUpIds":["gcp-compute","azure-vms"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"maturity + privacy + reliability.","tags":["enterprise"]},{"id":"managed-vector","slug":"managed-vector","name":"Managed vector database","shortName":"Managed vector database","problemStatement":"Zero-ops vector search for production RAG.","whenNotTopPick":"Already on Postgres <50M vectors — supabase-pgvector/Neon. Want absolute zero-ops brand — Pinecone.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"qdrant","depotPickRationale":"capability filtering + pricePerf + lockInRisk (prefer OSS escape).","runnersUpIds":["pinecone","weaviate","zilliz","upstash-vector","astra-db"],"weights":{"capabilityFit":0.24,"pricePerformance":0.18,"latencyReliability":0.18,"developerDx":0.16,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"capability filtering + pricePerf + lockInRisk (prefer OSS escape).","tags":["managed"]},{"id":"self-host-vector","slug":"self-host-vector","name":"Self-host vector DB","shortName":"Self-host vector DB","problemStatement":"Run vectors in your VPC/K8s with open source.","whenNotTopPick":"No ops appetite — Pinecone/Qdrant Cloud. Billion-scale specialized — Milvus/Zilliz.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"qdrant","depotPickRationale":"capability + lockInRisk low + dx of ops.","runnersUpIds":["weaviate","zilliz","chroma"],"weights":{"capabilityFit":0.24,"pricePerformance":0.18,"latencyReliability":0.18,"developerDx":0.16,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"capability + lockInRisk low + dx of ops.","tags":["self"]},{"id":"postgres-vector","slug":"postgres-vector","name":"Postgres-native vectors (pgvector)","shortName":"Postgres-native vectors (pgvector)","problemStatement":"Keep vectors next to relational data with minimal new systems.","whenNotTopPick":"Corpus/QPS exceeds comfortable Postgres vector range — Qdrant/Milvus/Turbopuffer.","categoryIds":["cat-vector-memory","cat-auth-payments"],"depotPickProviderId":"supabase-pgvector","depotPickRationale":"dx + pricePerf + lockInRisk. Default for most agent apps.","runnersUpIds":["neon","aws-ec2"],"weights":{"capabilityFit":0.24,"pricePerformance":0.18,"latencyReliability":0.18,"developerDx":0.16,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"dx + pricePerf + lockInRisk. Default for most agent apps.","tags":["postgres"]},{"id":"rag-memory","slug":"rag-memory","name":"Agent RAG memory store","shortName":"Agent RAG memory store","problemStatement":"Persistent retrieval memory for agents across sessions.","whenNotTopPick":"Hot session memory only — Redis. Huge cold corpora — Turbopuffer.","categoryIds":["cat-vector-memory","cat-auth-payments"],"depotPickProviderId":"supabase-pgvector","depotPickRationale":"balance of capability and operational simplicity.","runnersUpIds":["qdrant","turbopuffer","pinecone","redis","upstash-redis","upstash-vector"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"balance of capability and operational simplicity.","tags":["rag"]},{"id":"cheap-vector-scale","slug":"cheap-vector-scale","name":"Cost-efficient large vector corpora","shortName":"Cost-efficient large vector corpora","problemStatement":"Store/search very large or cold-heavy vector sets without Pinecone-scale bills.","whenNotTopPick":"Tiny corpora — pgvector. Need fancy hybrid modules — Weaviate.","categoryIds":["cat-vector-memory","cat-object-storage"],"depotPickProviderId":"turbopuffer","depotPickRationale":"pricePerf at scale + capability.","runnersUpIds":["qdrant","zilliz","supabase-pgvector","upstash-vector"],"weights":{"capabilityFit":0.24,"pricePerformance":0.18,"latencyReliability":0.18,"developerDx":0.16,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"pricePerf at scale + capability.","tags":["cheap"]},{"id":"billion-scale-vector","slug":"billion-scale-vector","name":"Billion-scale vector search","shortName":"Billion-scale vector search","problemStatement":"ANN at enormous scale with serious indexing.","whenNotTopPick":"Under ~50–100M vectors — don't pay the complexity tax.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"zilliz","depotPickRationale":"capability scale + maturity.","runnersUpIds":["qdrant","pinecone","turbopuffer"],"weights":{"capabilityFit":0.24,"pricePerformance":0.18,"latencyReliability":0.18,"developerDx":0.16,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"capability scale + maturity.","tags":["billion"]},{"id":"prototype-rag","slug":"prototype-rag","name":"Prototype RAG in a day","shortName":"Prototype RAG in a day","problemStatement":"Hackathon/demo RAG with minimal setup.","whenNotTopPick":"Production traffic — graduate immediately.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"chroma","depotPickRationale":"dx first; reliability discounted.","runnersUpIds":["supabase-pgvector","pinecone","upstash-vector"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx first; reliability discounted.","tags":["prototype"]},{"id":"agent-memory-kv","slug":"agent-memory-kv","name":"Hot agent KV / session memory","shortName":"Hot agent KV / session memory","problemStatement":"Sub-ms session state, caches, and short-term memory.","whenNotTopPick":"Durable relational memory — Postgres. Edge-only light state — Durable Objects.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"redis","depotPickRationale":"reliability + latency capability.","runnersUpIds":["supabase-pgvector","cloudflare-workers","upstash-redis"],"weights":{"capabilityFit":0.2,"pricePerformance":0.15,"latencyReliability":0.3,"developerDx":0.15,"privacyPosture":0.08,"lockInRisk":0.05,"maturity":0.07},"methodologyNotes":"reliability + latency capability.","tags":["agent"]},{"id":"llm-tracing","slug":"llm-tracing","name":"LLM/agent tracing","shortName":"LLM/agent tracing","problemStatement":"See prompts, tool calls, latency, and failures in production.","whenNotTopPick":"Deep LangChain shop — LangSmith. Want proxy in 10 minutes — Helicone. OSS-only — Phoenix. Python/PydanticAI traces + app OTel together — Logfire. Free eval+trace SKU — HoneyHive.","categoryIds":["cat-observability"],"depotPickProviderId":"langfuse","depotPickRationale":"dx + pricePerf + privacy (self-host option).","runnersUpIds":["helicone","langsmith","arize-phoenix","portkey","langwatch","comet-opik","logfire","honeyhive"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"dx + pricePerf + privacy (self-host option).","tags":["llm"]},{"id":"evals","slug":"evals","name":"Prompt/agent eval harness","shortName":"Prompt/agent eval harness","problemStatement":"Regression-test agent quality in CI before ship.","whenNotTopPick":"Budget OSS-only — Langfuse/Phoenix. Already standardized on LangSmith datasets. Humanloop is sunset (Anthropic acquihire) — do not pick.","categoryIds":["cat-observability"],"depotPickProviderId":"braintrust","depotPickRationale":"capability eval workflows + dx.","runnersUpIds":["langfuse","langsmith","arize-phoenix","confident-ai","langwatch","comet-opik","honeyhive","logfire"],"weights":{"capabilityFit":0.26,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.22,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"capability eval workflows + dx.","tags":["evals"]},{"id":"cost-monitoring","slug":"cost-monitoring","name":"LLM cost monitoring","shortName":"LLM cost monitoring","problemStatement":"Track token spend per user/feature and catch bill spikes.","whenNotTopPick":"Need deep evals more than cost — Braintrust; need unified cloud+GPU+LLM bills not just token traces — Vantage; want a no-SDK drop-in proxy and accept maintenance-mode risk (since 2026-03-03, ownership unclear) — Helicone.","categoryIds":["cat-observability"],"depotPickProviderId":"langfuse","depotPickRationale":"Editorial replacement 2026-10-21: the former pick Helicone (8.13 for this job) has been in maintenance mode since 2026-03-03 (https://www.mintlify.com/blog/mintlify-acquires-helicone: 'Helicone will continue operating in maintenance mode') and its site now says 'Helicone is no longer part of Mintlify. Details to come soon.' (https://www.helicone.ai/, re-checked 2026-10-21), so ownership and roadmap are unclear. Replaced by the highest-scoring live runner-up, Langfuse (8.66 for this job; cost tracking per model/user/feature, spend-threshold alerts and a Metrics API per https://langfuse.com/docs/observability/features/token-and-cost-tracking, re-checked 2026-10-21). Helicone stays a runner-up with the maintenance_mode flag. No score changed.","runnersUpIds":["openrouter","vantage","portkey","helicone"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + pricePerf of the observability tool itself.","tags":["cost"]},{"id":"experiment-tracking","slug":"experiment-tracking","name":"ML experiment tracking","shortName":"ML experiment tracking","problemStatement":"Track training runs, metrics, artifacts for fine-tunes.","whenNotTopPick":"Pure LLM app tracing without training — Langfuse.","categoryIds":["cat-observability","cat-train-finetune"],"depotPickProviderId":"weights-biases","depotPickRationale":"maturity + capability for training workflows.","runnersUpIds":["arize-phoenix","langfuse"],"weights":{"capabilityFit":0.26,"pricePerformance":0.22,"latencyReliability":0.1,"developerDx":0.18,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"maturity + capability for training workflows.","tags":["experiment"]},{"id":"infra-observability","slug":"infra-observability","name":"Infra-wide agent observability","shortName":"Infra-wide agent observability","problemStatement":"Debug distributed agent systems beyond prompts (queues, gateways, VMs).","whenNotTopPick":"Only need prompt traces — Langfuse/Helicone.","categoryIds":["cat-observability"],"depotPickProviderId":"honeycomb","depotPickRationale":"capability high-cardinality + reliability.","runnersUpIds":["arize-phoenix","langfuse"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability high-cardinality + reliability.","tags":["infra"]},{"id":"cloud-browser","slug":"cloud-browser","name":"Cloud browser for agents","shortName":"Cloud browser for agents","problemStatement":"Give agents Playwright-grade browsers with stealth/sessions managed.","whenNotTopPick":"Must self-host — Steel/Browserless. Only need static scrape — Firecrawl.","categoryIds":["cat-sandboxes-browsers","cat-search-scrape"],"depotPickProviderId":"browserbase","depotPickRationale":"dx + capability agent features + reliability.","runnersUpIds":["steel-dev","browserless","cloudflare-browser-rendering"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + capability agent features + reliability.","tags":["cloud"]},{"id":"self-host-browser","slug":"self-host-browser","name":"Self-host browser automation","shortName":"Self-host browser automation","problemStatement":"Run Chromium fleets in your infra.","whenNotTopPick":"No ops — Browserbase.","categoryIds":["cat-sandboxes-browsers","cat-search-scrape"],"depotPickProviderId":"steel-dev","depotPickRationale":"lockInRisk low + capability.","runnersUpIds":["browserless","playwright","puppeteer"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"lockInRisk low + capability.","tags":["self"]},{"id":"web-scrape","slug":"web-scrape","name":"Web scraping / page fetch","shortName":"Web scraping / page fetch","problemStatement":"Fetch and clean web content for agents.","whenNotTopPick":"Hard anti-bot enterprise — Bright Data. Interactive multi-step — Browserbase.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"firecrawl","depotPickRationale":"dx for LLM-ready output + capability.","runnersUpIds":["scrapingbee","bright-data","browserbase","scrapfly","zenrows","apify","jina","diffbot"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx for LLM-ready output + capability.","tags":["web"]},{"id":"llm-ready-extract","slug":"llm-ready-extract","name":"LLM-ready web extraction","shortName":"LLM-ready web extraction","problemStatement":"Turn sites into markdown/structured JSON for RAG.","whenNotTopPick":"Need search only without crawl — Exa/Tavily.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"firecrawl","depotPickRationale":"capability extraction quality + dx.","runnersUpIds":["scrapingbee","exa","scrapfly","zenrows","agentql","jina","diffbot"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability extraction quality + dx.","tags":["llm"]},{"id":"serp-scrape","slug":"serp-scrape","name":"SERP / search results API","shortName":"SERP / search results API","problemStatement":"Programmatic search engine results pages.","whenNotTopPick":"privacy index — Brave; AI-agent search — Tavily/Exa.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"serper","depotPickRationale":"pricePerf + capability; compliance caveats for scrape-based SERP.","runnersUpIds":["bright-data","brave-search","serpapi"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf + capability; compliance caveats for scrape-based SERP.","tags":["serp"]},{"id":"enterprise-proxy","slug":"enterprise-proxy","name":"Enterprise proxy & unlocker network","shortName":"Enterprise proxy & unlocker network","problemStatement":"Bypass tough anti-bot at scale with compliant enterprise tooling.","whenNotTopPick":"Simple pages — Firecrawl. Ethical/ToS constraints — don't.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"bright-data","depotPickRationale":"capability network + maturity; pricePerf low.","runnersUpIds":["scrapingbee","browserbase","scrapfly","zenrows","oxylabs"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability network + maturity; pricePerf low.","tags":["enterprise"]},{"id":"web-agent","slug":"web-agent","name":"Autonomous web-acting agent","shortName":"Autonomous web-acting agent","problemStatement":"Agent logs in, clicks, fills forms, completes workflows in browsers.","whenNotTopPick":"Read-only extraction — Firecrawl. Pure search — Exa/Tavily.","categoryIds":["cat-sandboxes-browsers","cat-search-scrape"],"depotPickProviderId":"browserbase","depotPickRationale":"capability browser + dx agent SDKs.","runnersUpIds":["steel-dev","browserless","firecrawl"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability browser + dx agent SDKs.","tags":["web"]},{"id":"agent-sandbox","slug":"agent-sandbox","name":"Secure code sandbox for agents","shortName":"Secure code sandbox for agents","problemStatement":"Execute untrusted LLM-generated code safely.","whenNotTopPick":"Need persistent full IDE workspaces — Daytona. Python jobs that aren't untrusted REPL — Modal.","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"e2b","depotPickRationale":"privacy/isolation capability + dx + reliability.","runnersUpIds":["modal","daytona","fly-machines","upstash-box"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"privacy/isolation capability + dx + reliability.","tags":["agent"]},{"id":"code-interpreter","slug":"code-interpreter","name":"Code interpreter tool for LLMs","shortName":"Code interpreter tool for LLMs","problemStatement":"Give any model a sandboxed REPL without OpenAI lock-in.","whenNotTopPick":"Happy inside one vendor's hosted interpreter and accept lock-in.","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"e2b","depotPickRationale":"dx + lockInRisk (prefer E2B) + isolation.","runnersUpIds":["modal","anthropic","openai","upstash-box"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + lockInRisk (prefer E2B) + isolation.","tags":["code"]},{"id":"persistent-dev-env","slug":"persistent-dev-env","name":"Persistent agent dev environments","shortName":"Persistent agent dev environments","problemStatement":"Agents need sticky workspaces with files/git across sessions.","whenNotTopPick":"Disposable single-run code — E2B.","categoryIds":["cat-sandboxes-browsers","cat-cpu-compute"],"depotPickProviderId":"daytona","depotPickRationale":"capability persistence + dx.","runnersUpIds":["e2b","fly-machines","railway"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability persistence + dx.","tags":["persistent"]},{"id":"transactional-email","slug":"transactional-email","name":"Transactional email","shortName":"Transactional email","problemStatement":"Send receipts, magic links, agent notifications reliably.","whenNotTopPick":"Maximum deliverability paranoia — Postmark. Huge legacy volume — SendGrid.","categoryIds":["cat-comms"],"depotPickProviderId":"resend","depotPickRationale":"dx + reliability; Postmark favored when deliverability is the crisis.","runnersUpIds":["postmark","sendgrid","customer-io","amazon-ses","mailgun"],"weights":{"capabilityFit":0.2,"pricePerformance":0.18,"latencyReliability":0.22,"developerDx":0.2,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.06},"methodologyNotes":"dx + reliability; Postmark favored when deliverability is the crisis.","tags":["transactional"]},{"id":"marketing-email","slug":"marketing-email","name":"Marketing / blast email","shortName":"Marketing / blast email","problemStatement":"Newsletters and campaigns (not pure transactional).","whenNotTopPick":"Transactional-only — Postmark/Resend.","categoryIds":["cat-comms"],"depotPickProviderId":"sendgrid","depotPickRationale":"capability marketing features + maturity.","runnersUpIds":["loops","resend","courier","customer-io"],"weights":{"capabilityFit":0.2,"pricePerformance":0.18,"latencyReliability":0.22,"developerDx":0.2,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.06},"methodologyNotes":"capability marketing features + maturity.","tags":["marketing"]},{"id":"sms","slug":"sms","name":"Programmable SMS","shortName":"Programmable SMS","problemStatement":"Send/receive SMS for agents and auth.","whenNotTopPick":"Email-only notifications — Resend. Multi-channel orchestration — Courier on top of Twilio.","categoryIds":["cat-comms"],"depotPickProviderId":"twilio","depotPickRationale":"capability coverage + maturity.","runnersUpIds":["courier","bird","onesignal","telnyx","plivo"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability coverage + maturity.","tags":["sms"]},{"id":"voice","slug":"voice","name":"Programmable voice / phone agents","shortName":"Programmable voice / phone agents","problemStatement":"Inbound/outbound voice connected to LLM agents.","whenNotTopPick":"TTS-only without telephony — ElevenLabs.","categoryIds":["cat-comms"],"depotPickProviderId":"twilio","depotPickRationale":"capability telephony + reliability; pair with ElevenLabs for voice quality.","runnersUpIds":["elevenlabs","telnyx","plivo","rime"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability telephony + reliability; pair with ElevenLabs for voice quality.","tags":["voice"]},{"id":"whatsapp","slug":"whatsapp","name":"WhatsApp messaging","shortName":"WhatsApp messaging","problemStatement":"Agent messaging over WhatsApp Business APIs.","whenNotTopPick":"Not needed outside supported regions/use cases.","categoryIds":["cat-comms"],"depotPickProviderId":"twilio","depotPickRationale":"capability + maturity of WhatsApp channels.","runnersUpIds":["courier","bird","telnyx","plivo"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability + maturity of WhatsApp channels.","tags":["whatsapp"]},{"id":"notification-orchestration","slug":"notification-orchestration","name":"Multi-channel notification orchestration","shortName":"Multi-channel notification orchestration","problemStatement":"Route email/SMS/push/Slack with preferences and templates.","whenNotTopPick":"Single channel — talk to Resend/Twilio directly.","categoryIds":["cat-comms"],"depotPickProviderId":"courier","depotPickRationale":"dx + capability orchestration.","runnersUpIds":["novu","knock","twilio","resend","onesignal","ably","customer-io","bird","svix"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + capability orchestration.","tags":["notification"]},{"id":"auth","slug":"auth","name":"Product authentication","shortName":"Product authentication","problemStatement":"User signup/login/organizations for an agent product.","whenNotTopPick":"Enterprise SSO-first motion — WorkOS. Full IdP customization — Auth0. Want auth+DB one vendor — Supabase.","categoryIds":["cat-auth-payments"],"depotPickProviderId":"clerk","depotPickRationale":"dx + capability orgs + maturity.","runnersUpIds":["auth0","supabase-pgvector","workos","stytch"],"weights":{"capabilityFit":0.24,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.24,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"dx + capability orgs + maturity.","tags":["auth"]},{"id":"enterprise-sso","slug":"enterprise-sso","name":"Enterprise SSO / SCIM","shortName":"Enterprise SSO / SCIM","problemStatement":"Close enterprise deals that require Okta/Azure AD SSO and directory sync.","whenNotTopPick":"Consumer-only app — Clerk without enterprise SSO yet.","categoryIds":["cat-auth-payments"],"depotPickProviderId":"workos","depotPickRationale":"capability SSO/SCIM + dx + privacy.","runnersUpIds":["auth0","clerk","stytch"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"capability SSO/SCIM + dx + privacy.","tags":["enterprise"]},{"id":"user-management","slug":"user-management","name":"User management UI & orgs","shortName":"User management UI & orgs","problemStatement":"Prebuilt user profiles, orgs, and session management.","whenNotTopPick":"Custom build on Supabase Auth / Cognito for cost control at massive MAU.","categoryIds":["cat-auth-payments"],"depotPickProviderId":"clerk","depotPickRationale":"dx primary.","runnersUpIds":["auth0","supabase-pgvector","stytch"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx primary.","tags":["user"]},{"id":"payments","slug":"payments","name":"Payments acceptance","shortName":"Payments acceptance","problemStatement":"Charge customers for the agent product.","whenNotTopPick":"Regions/business models Stripe can't support — then local PSPs (out of seed scope).","categoryIds":["cat-auth-payments"],"depotPickProviderId":"stripe","depotPickRationale":"maturity + dx + capability. De facto default.","runnersUpIds":["polar"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"maturity + dx + capability. De facto default.","tags":["payments"]},{"id":"billing","slug":"billing","name":"Subscription & usage billing","shortName":"Subscription & usage billing","problemStatement":"Seat + metered token/API billing for AI products.","whenNotTopPick":"Complex AI pricing simulations — Orb; OSS self-host metering — Lago; API gateway entitlements/token meters — OpenMeter; very custom ERP — still usually sync from Stripe.","categoryIds":["cat-auth-payments"],"depotPickProviderId":"stripe","depotPickRationale":"Stripe remains cash collection default; Orb/Lago/OpenMeter for serious usage meters.","runnersUpIds":["orb","lago","openmeter","metronome","autumn","polar"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Stripe remains cash collection default; Orb/Lago/OpenMeter for serious usage meters.","tags":["billing"]},{"id":"image-generation","slug":"image-generation","name":"Programmatic image generation","shortName":"Programmatic image generation","problemStatement":"Generate images from agents/apps via API.","whenNotTopPick":"Need Midjourney aesthetics and have API access; want OpenAI one-vendor stack; first-party FLUX.2/FLUX 3 terms — Black Forest Labs.","categoryIds":["cat-media-gen","cat-inference-apis"],"depotPickProviderId":"fal","depotPickRationale":"dx + capability catalog + pricePerf. Midjourney penalized on API dx.","runnersUpIds":["openai-media","black-forest-labs","stability","replicate","midjourney","ideogram","recraft"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + capability catalog + pricePerf. Midjourney penalized on API dx.","tags":["image"]},{"id":"video-generation","slug":"video-generation","name":"Programmatic video generation","shortName":"Programmatic video generation","problemStatement":"Generate short videos via API for products/agents.","whenNotTopPick":"Creative suite workflows for humans — Runway UI; confirm Pika API maturity.","categoryIds":["cat-media-gen","cat-inference-apis"],"depotPickProviderId":"fal","depotPickRationale":"capability quality + dx API + pricePerf.","runnersUpIds":["runway","luma","pika","minimax","kling"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability quality + dx API + pricePerf.","tags":["video"]},{"id":"media-generation","slug":"media-generation","name":"Unified media generation API","shortName":"Unified media generation API","problemStatement":"One API across many image/video/audio models.","whenNotTopPick":"Single-model specialist needs only.","categoryIds":["cat-media-gen","cat-inference-apis"],"depotPickProviderId":"fal","depotPickRationale":"dx + catalog capability.","runnersUpIds":["replicate","openai-media","black-forest-labs","ideogram","recraft"],"weights":{"capabilityFit":0.3,"pricePerformance":0.18,"latencyReliability":0.14,"developerDx":0.16,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"dx + catalog capability.","tags":["media"]},{"id":"speech-synthesis","slug":"speech-synthesis","name":"Text-to-speech","shortName":"Text-to-speech","problemStatement":"Natural TTS for agents and products.","whenNotTopPick":"Already all-in on OpenAI and quality sufficient.","categoryIds":["cat-media-gen","cat-comms"],"depotPickProviderId":"elevenlabs","depotPickRationale":"capability voice quality + dx.","runnersUpIds":["cartesia","openai-media","hume","minimax","rime"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability voice quality + dx.","tags":["speech"]},{"id":"speech-to-text","slug":"speech-to-text","name":"Speech-to-text / transcription","shortName":"Speech-to-text / transcription","problemStatement":"Transcribe audio for agents.","whenNotTopPick":"Already all-in on OpenAI and Whisper quality/latency is fine — openai-media; need richest audio-intelligence features (summaries/PII) — AssemblyAI; ultra-cheap batch Whisper on Groq LPU — groq.","categoryIds":["cat-media-gen","cat-inference-apis"],"depotPickProviderId":"deepgram","depotPickRationale":"Blunt: openai-media was the Round-1 convenience pick before Deepgram entered the catalog. Deepgram wins production STT on $/min + streaming agent DX (Nova-3 / Flux). Keep openai-media only when you refuse a second vendor. Aligns with production-stt.","runnersUpIds":["assemblyai","openai-media","groq","speechmatics","gladia"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Blunt: openai-media was the Round-1 convenience pick before Deepgram entered the catalog. Deepgram wins production STT on $/min + streaming agent DX (Nova-3 / Flux). Keep openai-media only when you refuse a second vendor. Aligns with production-stt.","tags":["speech"]},{"id":"voice-agents","slug":"voice-agents","name":"Voice agents (talking products)","shortName":"Voice agents (talking products)","problemStatement":"Build agents that speak and listen on phone/web.","whenNotTopPick":"Telephony-only plumbing without premium TTS — Twilio + cheaper TTS.","categoryIds":["cat-media-gen","cat-comms"],"depotPickProviderId":"elevenlabs","depotPickRationale":"capability voice agents + integrations.","runnersUpIds":["twilio","openai","retell","bland"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability voice agents + integrations.","tags":["voice"]},{"id":"3d-generation","slug":"3d-generation","name":"3D generation","shortName":"3D generation","problemStatement":"Generate 3D assets from prompts/images.","whenNotTopPick":"2D-only pipelines.","categoryIds":["cat-media-gen"],"depotPickProviderId":"luma","depotPickRationale":"capability 3D + dx API.","runnersUpIds":["fal","replicate"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability 3D + dx API.","tags":["3d"]},{"id":"agent-search","slug":"agent-search","name":"Web search tool for agents","shortName":"Web search tool for agents","problemStatement":"Give agents a search tool that returns clean results for reasoning.","whenNotTopPick":"Keyword SERP style preferred — Tavily/Serper; need Task/deep-research in one API — Parallel; privacy — Brave.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"exa","depotPickRationale":"capability for agent payloads + dx.","runnersUpIds":["tavily","parallel","brave-search","serper","serpapi","you-com","linkup"],"weights":{"capabilityFit":0.28,"pricePerformance":0.16,"latencyReliability":0.14,"developerDx":0.18,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"capability for agent payloads + dx.","tags":["agent"]},{"id":"neural-search","slug":"neural-search","name":"Neural / semantic web search","shortName":"Neural / semantic web search","problemStatement":"Find relevant pages by meaning for research agents.","whenNotTopPick":"Exact navigational queries — keyword SERP APIs.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"exa","depotPickRationale":"capability semantic retrieval.","runnersUpIds":["tavily","brave-search","typesense-cloud","algolia","meilisearch-cloud"],"weights":{"capabilityFit":0.26,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.22,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"capability semantic retrieval.","tags":["neural"]},{"id":"search-augmented-llm","slug":"search-augmented-llm","name":"Search-augmented generation","shortName":"Search-augmented generation","problemStatement":"Combine search + LLM for fresher answers.","whenNotTopPick":"Want fully managed cite-and-answer — Perplexity. Need full page text — Firecrawl after search.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"tavily","depotPickRationale":"dx agent integrations + capability.","runnersUpIds":["exa","perplexity","parallel","firecrawl","you-com"],"weights":{"capabilityFit":0.28,"pricePerformance":0.16,"latencyReliability":0.14,"developerDx":0.18,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"dx agent integrations + capability.","tags":["search"]},{"id":"privacy-search","slug":"privacy-search","name":"Privacy-oriented search API","shortName":"Privacy-oriented search API","problemStatement":"Search with independent index and stronger privacy posture.","whenNotTopPick":"Max recall on niche queries may still need multiple search providers.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"brave-search","depotPickRationale":"privacy + capability.","runnersUpIds":["exa","tavily"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"privacy + capability.","tags":["privacy"]},{"id":"enterprise-search","slug":"enterprise-search","name":"Enterprise web search API","shortName":"Enterprise web search API","problemStatement":"Official search API under enterprise cloud terms.","whenNotTopPick":"Indie DX priority — Tavily/Exa.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"brave-search","depotPickRationale":"Editorial replacement 2026-10-21: Microsoft retired the Bing Search APIs on 2025-08-11 (https://learn.microsoft.com/en-us/lifecycle/announcements/bing-search-api-retirement), so the former pick bing-search (8.06 for this job) was replaced by the highest-scoring live runner-up, Brave Search (8.41 for this job; pricing and status re-checked 2026-10-21). No score changed. Brave: own independent index, plan/credit-based API.","runnersUpIds":["tavily","algolia","typesense-cloud","meilisearch-cloud","elastic-serverless"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.12,"privacyPosture":0.25,"lockInRisk":0.09,"maturity":0.1},"methodologyNotes":"maturity + privacy/compliance + reliability.","tags":["enterprise"]},{"id":"object-storage","slug":"object-storage","name":"Object storage for agent artifacts","shortName":"Object storage for agent artifacts","problemStatement":"Store files, datasets, and agent outputs.","whenNotTopPick":"Deep AWS estate / S3-only features — S3. Global $0-egress S3 without Cloudflare — Tigris. Cheapest always-hot TB — Backblaze B2. Self-host/VPC — MinIO. Want Storage in BaaS — Supabase.","categoryIds":["cat-object-storage"],"depotPickProviderId":"r2","depotPickRationale":"pricePerf (egress) + compatibility.","runnersUpIds":["s3","tigris","backblaze-b2","digitalocean","minio","wasabi","google-cloud-storage","azure-blob"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf (egress) + compatibility.","tags":["object"]},{"id":"cheap-egress-storage","slug":"cheap-egress-storage","name":"Object storage without egress pain","shortName":"Object storage without egress pain","problemStatement":"Artifacts are downloaded often; S3 egress would dominate bill.","whenNotTopPick":"Need S3-specific ecosystem features unavailable on R2. Want global S3-compat $0 egress without CF — Tigris. OK with 3× free egress then $0.01/GB — B2. Must self-host — MinIO.","categoryIds":["cat-object-storage"],"depotPickProviderId":"r2","depotPickRationale":"pricePerf egress + compatibility.","runnersUpIds":["tigris","backblaze-b2","digitalocean","hetzner","minio","wasabi"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf egress + compatibility.","tags":["cheap"]},{"id":"serverless-db","slug":"serverless-db","name":"Serverless Postgres for agents","shortName":"Serverless Postgres for agents","problemStatement":"Scale-to-zero SQL with branching for previews/agents.","whenNotTopPick":"Want Auth+Storage+Realtime suite — Supabase.","categoryIds":["cat-vector-memory","cat-object-storage"],"depotPickProviderId":"neon","depotPickRationale":"dx + pricePerf + capability branching.","runnersUpIds":["supabase-pgvector","planetscale"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + pricePerf + capability branching.","tags":["serverless"]},{"id":"app-backend","slug":"app-backend","name":"All-in-one app backend (DB/auth/storage)","shortName":"All-in-one app backend (DB/auth/storage)","problemStatement":"Ship agent product backend without stitching five vendors.","whenNotTopPick":"Best-of-breed auth (Clerk) + DB (Neon) preferred for scale/flexibility.","categoryIds":["cat-vector-memory","cat-auth-payments"],"depotPickProviderId":"supabase-pgvector","depotPickRationale":"dx + integration capability.","runnersUpIds":["neon","railway","clerk","planetscale"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + integration capability.","tags":["app"]},{"id":"hybrid-search-rag","slug":"hybrid-search-rag","name":"Hybrid keyword + vector RAG","shortName":"Hybrid keyword + vector RAG","problemStatement":"Need BM25/keyword + vector in one system.","whenNotTopPick":"Simple semantic-only — Qdrant/pgvector.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"weaviate","depotPickRationale":"capability hybrid search.","runnersUpIds":["qdrant","supabase-pgvector","zilliz","typesense-cloud","algolia","meilisearch-cloud","elastic-serverless"],"weights":{"capabilityFit":0.28,"pricePerformance":0.16,"latencyReliability":0.14,"developerDx":0.18,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"capability hybrid search.","tags":["hybrid"]},{"id":"budget-open-inference","slug":"budget-open-inference","name":"Ultra-budget open-model inference","shortName":"Ultra-budget open-model inference","problemStatement":"Need the cheapest OpenAI-compatible host for high-volume open weights without caring about brand prestige.","whenNotTopPick":"Need Fireworks-level production features/tool calling reliability, or compliance blocks smaller hosts.","categoryIds":["cat-inference-apis","cat-gpu-cloud"],"depotPickProviderId":"deepinfra","depotPickRationale":"Overweight pricePerf. DeepInfra wins catalog+price; Novita/SiliconFlow/Hyperbolic as pure price peers.","runnersUpIds":["novita","siliconflow","hyperbolic","zai","chutes","minimax"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Overweight pricePerf. DeepInfra wins catalog+price; Novita/SiliconFlow/Hyperbolic as pure price peers.","tags":["budget"]},{"id":"edge-llm-inference","slug":"edge-llm-inference","name":"Edge-colocated LLM inference","shortName":"Edge-colocated LLM inference","problemStatement":"Agent runs on edge/Workers and needs low-latency small/open models next to the runtime.","whenNotTopPick":"Need frontier reasoning quality — use Anthropic/OpenAI and accept hop latency.","categoryIds":["cat-inference-apis","cat-cpu-compute"],"depotPickProviderId":"cloudflare-workers-ai","depotPickRationale":"Colocation + pricePerf. Workers AI for edge; Groq for raw speed off-edge.","runnersUpIds":["groq","fireworks","openrouter"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Colocation + pricePerf. Workers AI for edge; Groq for raw speed off-edge.","tags":["edge"]},{"id":"lakehouse-llm","slug":"lakehouse-llm","name":"LLM on existing lakehouse data","shortName":"LLM on existing lakehouse data","problemStatement":"Data already lives in Databricks or Snowflake; want governed inference/RAG without exporting corpora.","whenNotTopPick":"Greenfield agent with no lakehouse — use direct APIs + pgvector/Qdrant.","categoryIds":["cat-inference-apis","cat-train-finetune"],"depotPickProviderId":"databricks-mosaic","depotPickRationale":"Data gravity + privacy/governance beat raw DX.","runnersUpIds":["snowflake-cortex","amazon-bedrock","vertex-ai"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Data gravity + privacy/governance beat raw DX.","tags":["lakehouse"]},{"id":"self-host-nim","slug":"self-host-nim","name":"Self-hosted optimized model serving","shortName":"Self-hosted optimized model serving","problemStatement":"Run optimized inference containers on your own NVIDIA GPUs across clouds.","whenNotTopPick":"Don't want to operate GPUs — use Fireworks/Together serverless.","categoryIds":["cat-inference-apis","cat-gpu-cloud"],"depotPickProviderId":"nvidia-nim","depotPickRationale":"Capability of CUDA stacks + privacy. Note: vllm-note is conceptual — prefer NIM or Baseten if no vLLM provider entry.","runnersUpIds":["baseten","fireworks","replicate"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability of CUDA stacks + privacy. Note: vllm-note is conceptual — prefer NIM or Baseten if no vLLM provider entry.","tags":["self"]},{"id":"neocloud-gpu","slug":"neocloud-gpu","name":"Neocloud reserved / list-price GPUs","shortName":"Neocloud reserved / list-price GPUs","problemStatement":"Need predictable H100/H200/Blackwell capacity with public list prices and cluster features.","whenNotTopPick":"Need cheapest interruptible — Vast/Salad/TensorDock.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"nebius","depotPickRationale":"Reliability + transparent pricePerf. Nebius list prices are unusually clear.","runnersUpIds":["lambda","coreweave","massed-compute","verda","hyperstack","gcore","cloudferro","koyeb"],"weights":{"capabilityFit":0.22,"pricePerformance":0.25,"latencyReliability":0.15,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Reliability + transparent pricePerf. Nebius list prices are unusually clear.","tags":["neocloud"]},{"id":"distributed-cheap-gpu","slug":"distributed-cheap-gpu","name":"Ultra-cheap interruptible GPUs","shortName":"Ultra-cheap interruptible GPUs","problemStatement":"Batch/vision jobs that can restart; minimize $/GPU-hr on consumer or spot capacity.","whenNotTopPick":"Need SLA, multi-node IB, or customer PII guarantees.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"salad","depotPickRationale":"Pure pricePerf; reliability intentionally discounted.","runnersUpIds":["vast-ai","tensordock","oblivus"],"weights":{"capabilityFit":0.22,"pricePerformance":0.25,"latencyReliability":0.15,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Pure pricePerf; reliability intentionally discounted.","tags":["distributed"]},{"id":"gpu-marketplace-router","slug":"gpu-marketplace-router","name":"Multi-cloud GPU shopping / routing","shortName":"Multi-cloud GPU shopping / routing","problemStatement":"Compare and provision GPUs across many clouds from one API without N consoles.","whenNotTopPick":"Already standardized on one neocloud with commits.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"shadeform","depotPickRationale":"DX of aggregation; remember underlying provider SLA still binds.","runnersUpIds":["vast-ai","runpod","tensordock"],"weights":{"capabilityFit":0.22,"pricePerformance":0.25,"latencyReliability":0.15,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"DX of aggregation; remember underlying provider SLA still binds.","tags":["gpu"]},{"id":"agent-tool-router","slug":"agent-tool-router","name":"SaaS tool / MCP router for agents","shortName":"SaaS tool / MCP router for agents","problemStatement":"Connect agents to many third-party apps (Gmail, GitHub, CRMs) with managed OAuth/MCP.","whenNotTopPick":"Only need browsers or one API — call providers directly.","categoryIds":["cat-sandboxes-browsers","cat-search-scrape"],"depotPickProviderId":"composio","depotPickRationale":"DX + capability of tool catalog; watch lock-in.","runnersUpIds":["browserbase","anchor-browser","steel-dev"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"DX + capability of tool catalog; watch lock-in.","tags":["agent"]},{"id":"computer-use-browser","slug":"computer-use-browser","name":"Computer-use / stealth cloud browsers","shortName":"Computer-use / stealth cloud browsers","problemStatement":"Agents must operate authenticated, bot-hostile sites with human-like browsers.","whenNotTopPick":"Simple scrape/extract — Firecrawl; self-host — Steel; hosted CUA models on stealth Chromium — Hyperbrowser; vision RPA workflows — Skyvern.","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"anchor-browser","depotPickRationale":"Capability on stealth/auth + DX (MCP/SDK).","runnersUpIds":["browserbase","kernel","hyperbrowser","browser-use"],"weights":{"capabilityFit":0.24,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.24,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"Capability on stealth/auth + DX (MCP/SDK).","tags":["computer"]},{"id":"browser-agent-loop","slug":"browser-agent-loop","name":"Managed browser-agent loop","shortName":"Managed browser-agent loop","problemStatement":"Want NL goal → browser agent with optional raw CDP, not just a session host.","whenNotTopPick":"You already have an agent framework and only need hosted Chromium.","categoryIds":["cat-sandboxes-browsers","cat-search-scrape"],"depotPickProviderId":"browser-use","depotPickRationale":"Agent-loop capability over raw browser hosting.","runnersUpIds":["browserbase","anchor-browser","steel-dev"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Agent-loop capability over raw browser hosting.","tags":["browser"]},{"id":"platform-sandbox","slug":"platform-sandbox","name":"Sandbox + full app platform","shortName":"Sandbox + full app platform","problemStatement":"Agents need isolated compute plus APIs, DBs, GPUs, or BYOC — not only a code jail.","whenNotTopPick":"Disposable code exec only — E2B; simple web app — Railway.","categoryIds":["cat-cpu-compute","cat-sandboxes-browsers"],"depotPickProviderId":"northflank","depotPickRationale":"Capability breadth of platform vs pure sandbox DX.","runnersUpIds":["daytona","modal","railway"],"weights":{"capabilityFit":0.3,"pricePerformance":0.15,"latencyReliability":0.15,"developerDx":0.2,"privacyPosture":0.08,"lockInRisk":0.05,"maturity":0.07},"methodologyNotes":"Capability breadth of platform vs pure sandbox DX.","tags":["platform"]},{"id":"hosted-agent-memory","slug":"hosted-agent-memory","name":"Hosted agent memory layer","shortName":"Hosted agent memory layer","problemStatement":"Bolt persistent memory (add/retrieve) onto any agent without building a vector stack.","whenNotTopPick":"Need temporal knowledge graphs — Zep; stateful agent runtime — Letta; already have pgvector.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"mem0","depotPickRationale":"DX + pricePerf for bolt-on memory; watch retrieval metering.","runnersUpIds":["zep","letta","supabase-pgvector","supermemory","memobase","cognee"],"weights":{"capabilityFit":0.26,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.22,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"DX + pricePerf for bolt-on memory; watch retrieval metering.","tags":["hosted"]},{"id":"temporal-graph-memory","slug":"temporal-graph-memory","name":"Temporal / graph agent memory","shortName":"Temporal / graph agent memory","problemStatement":"Agents must recall entities, relationships, and time-evolving facts — not just similar chunks.","whenNotTopPick":"Simple preference/memory snippets — Mem0; pure vector RAG — Qdrant/pgvector.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"zep","depotPickRationale":"Capability of temporal graph memory over raw vector recall.","runnersUpIds":["mem0","weaviate","letta","cognee"],"weights":{"capabilityFit":0.24,"pricePerformance":0.18,"latencyReliability":0.18,"developerDx":0.16,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"Capability of temporal graph memory over raw vector recall.","tags":["temporal"]},{"id":"stateful-agent-runtime","slug":"stateful-agent-runtime","name":"Stateful long-lived agent runtime","shortName":"Stateful long-lived agent runtime","problemStatement":"Need agents with editable memory blocks and long-lived identity, not just a chat API + DB.","whenNotTopPick":"Stateless request/response agents — plain LLM API + Mem0.","categoryIds":["cat-vector-memory","cat-sandboxes-browsers"],"depotPickProviderId":"letta","depotPickRationale":"Runtime capability; Letta for memory-native agents, LangGraph/Temporal for orchestration.","runnersUpIds":["langgraph-cloud","temporal-cloud","mem0"],"weights":{"capabilityFit":0.26,"pricePerformance":0.16,"latencyReliability":0.16,"developerDx":0.18,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"Runtime capability; Letta for memory-native agents, LangGraph/Temporal for orchestration.","tags":["stateful"]},{"id":"embedded-vector","slug":"embedded-vector","name":"Embedded / object-store vector DB","shortName":"Embedded / object-store vector DB","problemStatement":"Want local or object-storage-centric vectors without a heavy managed cluster.","whenNotTopPick":"Need filtered multi-tenant production QPS — Qdrant/Pinecone.","categoryIds":["cat-vector-memory","cat-object-storage"],"depotPickProviderId":"lancedb","depotPickRationale":"pricePerf + DX for embedded; DuckDB for analytics-adjacent light vector.","runnersUpIds":["chroma","turbopuffer","duckdb"],"weights":{"capabilityFit":0.24,"pricePerformance":0.18,"latencyReliability":0.18,"developerDx":0.16,"privacyPosture":0.1,"lockInRisk":0.08,"maturity":0.06},"methodologyNotes":"pricePerf + DX for embedded; DuckDB for analytics-adjacent light vector.","tags":["embedded"]},{"id":"oss-evals","slug":"oss-evals","name":"OSS / CI-native LLM evals","shortName":"OSS / CI-native LLM evals","problemStatement":"Run prompt evals and red teams in CI without paying for a hosted eval SaaS.","whenNotTopPick":"Need hosted collaboration/datasets for PMs — Braintrust; RAG-only metrics — Ragas.","categoryIds":["cat-observability"],"depotPickProviderId":"promptfoo","depotPickRationale":"pricePerf + privacy of local runs; Promptfoo wins general OSS+redteam.","runnersUpIds":["ragas","braintrust","langsmith","confident-ai","comet-opik","langwatch"],"weights":{"capabilityFit":0.26,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.22,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"pricePerf + privacy of local runs; Promptfoo wins general OSS+redteam.","tags":["oss"]},{"id":"rag-evals","slug":"rag-evals","name":"RAG-specific evaluation metrics","shortName":"RAG-specific evaluation metrics","problemStatement":"Measure faithfulness, context precision/recall for retrieval pipelines.","whenNotTopPick":"General agent/tool evals — Promptfoo/Braintrust.","categoryIds":["cat-observability"],"depotPickProviderId":"ragas","depotPickRationale":"Capability fit for RAG metrics.","runnersUpIds":["promptfoo","braintrust","galileo"],"weights":{"capabilityFit":0.26,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.22,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"Capability fit for RAG metrics.","tags":["rag"]},{"id":"production-guardrail-evals","slug":"production-guardrail-evals","name":"Production guardrails & hallucination monitoring","shortName":"Production guardrails & hallucination mo…","problemStatement":"Detect and block bad outputs in production with research-grade metrics.","whenNotTopPick":"Pre-release simulation focus — Maxim; lightweight CI — Promptfoo.","categoryIds":["cat-observability"],"depotPickProviderId":"galileo","depotPickRationale":"Capability of guardrails + maturity.","runnersUpIds":["braintrust","maxim","langsmith","arthur-ai","patronus","confident-ai","honeyhive"],"weights":{"capabilityFit":0.26,"pricePerformance":0.12,"latencyReliability":0.12,"developerDx":0.22,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"Capability of guardrails + maturity.","tags":["production"]},{"id":"realtime-tts","slug":"realtime-tts","name":"Ultra-low-latency text-to-speech","shortName":"Ultra-low-latency text-to-speech","problemStatement":"Voice agents need streaming TTS with minimal time-to-first-audio.","whenNotTopPick":"Need largest voice library — ElevenLabs; already standardized on Deepgram Voice Agent; need expressive/empathic TTS — Hume. PlayHT/PlayAI is sunset (Meta) — do not pick.","categoryIds":["cat-media-gen"],"depotPickProviderId":"cartesia","depotPickRationale":"Latency capability + pricePerf.","runnersUpIds":["elevenlabs","deepgram","hume","rime"],"weights":{"capabilityFit":0.2,"pricePerformance":0.15,"latencyReliability":0.3,"developerDx":0.15,"privacyPosture":0.08,"lockInRisk":0.05,"maturity":0.07},"methodologyNotes":"Latency capability + pricePerf.","tags":["realtime"]},{"id":"production-stt","slug":"production-stt","name":"Production speech-to-text","shortName":"Production speech-to-text","problemStatement":"Accurate, cheap streaming and pre-recorded transcription for agents and voice apps.","whenNotTopPick":"Need richest audio-intelligence features in one API — AssemblyAI; only Whisper via OpenAI — openai-media.","categoryIds":["cat-media-gen","cat-inference-apis"],"depotPickProviderId":"deepgram","depotPickRationale":"pricePerf + capability; Deepgram rate card is the benchmark.","runnersUpIds":["assemblyai","openai-media","cartesia","speechmatics","gladia"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf + capability; Deepgram rate card is the benchmark.","tags":["production"]},{"id":"voice-agent-platform","slug":"voice-agent-platform","name":"Full voice-agent platform (STT+TTS+orchestration)","shortName":"Full voice-agent platform (STT+TTS+orche…","problemStatement":"Ship a realtime voice agent without stitching STT, LLM, and TTS yourself.","whenNotTopPick":"Want OSS realtime media control and to compose models yourself — LiveKit; STT-quality is the product — Deepgram; already on Cartesia Sonic+Ink and only need their voice agent minutes.","categoryIds":["cat-media-gen","cat-inference-apis"],"depotPickProviderId":"vapi","depotPickRationale":"End-to-end capability to ship without stitching. Round 4: Vapi is the actual orchestration SKU; Deepgram remains production-stt pick. Watch Deepgram Flux TTS promo expiry 2026-09-14.","runnersUpIds":["bland","retell","livekit","deepgram","cartesia"],"weights":{"capabilityFit":0.3,"pricePerformance":0.16,"latencyReliability":0.2,"developerDx":0.16,"privacyPosture":0.08,"lockInRisk":0.05,"maturity":0.05},"methodologyNotes":"End-to-end capability to ship without stitching. Round 4: Vapi is the actual orchestration SKU; Deepgram remains production-stt pick. Watch Deepgram Flux TTS promo expiry 2026-09-14.","tags":["voice"]},{"id":"durable-agent-workflows","slug":"durable-agent-workflows","name":"Durable long-running agent workflows","shortName":"Durable long-running agent workflows","problemStatement":"Multi-step agents that must survive crashes, retries, and human waits over hours/days.","whenNotTopPick":"Simple TS background jobs — Trigger.dev; event-driven steps with max DX — Inngest; already on LangGraph — LangGraph Platform.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"temporal-cloud","depotPickRationale":"Reliability first; Temporal for serious durability, Inngest for DX.","runnersUpIds":["inngest","trigger-dev","langgraph-cloud","hatchet","restate","dbos"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Reliability first; Temporal for serious durability, Inngest for DX.","tags":["durable"]},{"id":"event-driven-agent-jobs","slug":"event-driven-agent-jobs","name":"Event-driven agent jobs with great DX","shortName":"Event-driven agent jobs with great DX","problemStatement":"Fan-out, debounce, and retry agent work from events without operating queues.","whenNotTopPick":"Polyglot extreme scale — Temporal; raw HTTP delayed callbacks without a workflow runtime — Upstash QStash; pure Python GPU jobs — Modal.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"inngest","depotPickRationale":"DX + reliability balance; Inngest default for most product agents.","runnersUpIds":["trigger-dev","upstash-qstash","temporal-cloud","hatchet","restate","dbos"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"DX + reliability balance; Inngest default for most product agents.","tags":["event"]},{"id":"langgraph-deploy","slug":"langgraph-deploy","name":"Hosted LangGraph agent deployment","shortName":"Hosted LangGraph agent deployment","problemStatement":"Deploy LangGraph graphs with managed runtime and LangSmith traces.","whenNotTopPick":"Not using LangGraph — don't adopt just for hosting.","categoryIds":["cat-cpu-compute","cat-observability"],"depotPickProviderId":"langgraph-cloud","depotPickRationale":"Ecosystem fit; lock-in acknowledged.","runnersUpIds":["langsmith","inngest","temporal-cloud"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Ecosystem fit; lock-in acknowledged.","tags":["langgraph"]},{"id":"ray-distributed-agents","slug":"ray-distributed-agents","name":"Distributed Ray workloads for agents/ML","shortName":"Distributed Ray workloads for agents/ML","problemStatement":"Scale Python agent/ML workloads across a Ray cluster with managed control plane.","whenNotTopPick":"Single-endpoint serverless — Modal; lakehouse-native — Databricks.","categoryIds":["cat-gpu-cloud","cat-cpu-compute"],"depotPickProviderId":"anyscale","depotPickRationale":"Capability for distributed Python; Anyscale when Ray is the substrate.","runnersUpIds":["modal","coreweave","databricks-mosaic"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability for distributed Python; Anyscale when Ray is the substrate.","tags":["ray"]},{"id":"agent-secrets","slug":"agent-secrets","name":"Agent secrets & non-human credentials","shortName":"Agent secrets & non-human credentials","problemStatement":"Store and inject API keys for agents without leaking long-lived secrets into prompts, logs, or sandboxes.","whenNotTopPick":"Want seat-priced SaaS DX + free agent identities — Doppler; already standardized on 1Password Business — onepassword-secrets; HashiCorp estate / dynamic creds — hashicorp-vault; pure AWS IAM world — aws-secrets-manager.","categoryIds":["cat-auth-payments"],"depotPickProviderId":"infisical","depotPickRationale":"privacy + dx + lockInRisk. Infisical wins on Agent Proxy + MIT self-host; Doppler wins pure DX.","runnersUpIds":["doppler","onepassword-secrets","hashicorp-vault","aws-secrets-manager"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"privacy + dx + lockInRisk. Infisical wins on Agent Proxy + MIT self-host; Doppler wins pure DX.","tags":["agent"]},{"id":"content-moderation","slug":"content-moderation","name":"Content moderation (I/O safety classify)","shortName":"Content moderation (I/O safety classify)","problemStatement":"Classify user or model text/images for policy violations before acting.","whenNotTopPick":"Need open Llama Guard classifier with low latency — groq; need programmable multi-step rails — nvidia-nemo-guardrails; need EU-hosted guard models — ovhcloud-ai.","categoryIds":["cat-inference-apis","cat-observability"],"depotPickProviderId":"openai-moderation","depotPickRationale":"pricePerf + dx + reliability. Free OpenAI Moderation is the blunt default; escalate to Guardrails/Llama Guard for custom policy.","runnersUpIds":["groq","nvidia-nemo-guardrails","lakera","ovhcloud-ai","guardrails-ai"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf + dx + reliability. Free OpenAI Moderation is the blunt default; escalate to Guardrails/Llama Guard for custom policy.","tags":["content"]},{"id":"llm-input-guardrails","slug":"llm-input-guardrails","name":"Programmable LLM input/output guardrails","shortName":"Programmable LLM input/output guardrails","problemStatement":"Enforce custom safety, topical, and tool-use policies around agent LLM calls — not just a fixed taxonomy.","whenNotTopPick":"Need a dedicated prompt-injection API — lakera; only need free category flags — openai-moderation; production hallucination monitoring SaaS — galileo; cheap Llama Guard leaf — groq.","categoryIds":["cat-observability","cat-inference-apis"],"depotPickProviderId":"nvidia-nemo-guardrails","depotPickRationale":"capability of policy engine + privacy. Distinct from production-guardrail-evals (monitoring) — this is the runtime rails layer.","runnersUpIds":["lakera","galileo","openai-moderation","groq","guardrails-ai","arthur-ai","hiddenlayer"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability of policy engine + privacy. Distinct from production-guardrail-evals (monitoring) — this is the runtime rails layer.","tags":["llm"]},{"id":"usage-metering","slug":"usage-metering","name":"AI usage metering & entitlements","shortName":"AI usage metering & entitlements","problemStatement":"Meter tokens/tools/API calls and enforce entitlements for usage-based agent SaaS billing.","whenNotTopPick":"Simple subscription + basic meters — Stripe Billing; OSS self-host billing — Lago; heavy pricing experimentation/backtests — Orb.","categoryIds":["cat-auth-payments"],"depotPickProviderId":"openmeter","depotPickRationale":"capability for AI token meters + entitlements. OpenMeter/Kong for API-shaped products; Orb for pricing science; Lago for OSS.","runnersUpIds":["lago","orb","stripe","metronome","autumn","amberflo","polar"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability for AI token meters + entitlements. OpenMeter/Kong for API-shaped products; Orb for pricing science; Lago for OSS.","tags":["usage"]},{"id":"people-enrichment","slug":"people-enrichment","name":"People/company enrichment for research agents","shortName":"People/company enrichment for research agents","problemStatement":"Resolve emails, titles, and firmographics so research/GTM agents can act on real contacts.","whenNotTopPick":"Need full GTM sequencing + CRM-shaped workflow — Apollo.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"people-data-labs","depotPickRationale":"pricePerf + dx for automated pipelines. PDL for API-first agents; Apollo for sales-ops gravity.","runnersUpIds":["apollo","zoominfo","enrich-layer","clearbit-hubspot","parallel"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf + dx for automated pipelines. PDL for API-first agents; Apollo for sales-ops gravity.","tags":["people"]},{"id":"desktop-computer-use","slug":"desktop-computer-use","name":"Hosted desktop computer-use sandboxes","shortName":"Hosted desktop computer-use sandboxes","problemStatement":"Give agents a full GUI desktop (mouse/keyboard/screenshot), not just a headless browser.","whenNotTopPick":"Need Windows/macOS/Android fleets or local OSS sandboxes — Cua; need turnkey hosted Windows sessions — Scrapybara; browser-only stealth — Anchor/Hyperbrowser; real iOS/Android silicon — BrowserStack / AWS Device Farm (see mobile-computer-use).","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"e2b","depotPickRationale":"capability desktop APIs + isolation. E2B Desktop SDK is the programmable Linux default; Cua for multi-OS fleets; Scrapybara for turnkey hosted desktops.","runnersUpIds":["scrapybara","cua","anchor-browser"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability desktop APIs + isolation. E2B Desktop SDK is the programmable Linux default; Cua for multi-OS fleets; Scrapybara for turnkey hosted desktops.","tags":["desktop"]},{"id":"mcp-discovery","slug":"mcp-discovery","name":"MCP server discovery & registries","shortName":"MCP server discovery & registries","problemStatement":"Find, verify, and optionally host MCP tool servers for agents.","whenNotTopPick":"Want a directory/inspector on top of registry metadata — Glama or PulseMCP; a public directory without connect/API (after 2026-10-31) — Smithery; need broad SaaS tool router not MCP-native — Composio.","categoryIds":["cat-search-scrape","cat-sandboxes-browsers"],"depotPickProviderId":"mcp-registry","depotPickRationale":"Editorial replacement 2026-10-24: Smithery's own page (https://smithery.ai/future, fetched 2026-10-24) says 'On October 31, 2026, we're shutting down Smithery's connection management services.' and that the Smithery API, Connect, accounts and API keys, CLI and SDKs stop working, while smithery.ai stays a public (non-proxied) directory. So the former pick smithery (7.7 for this job) was replaced by the highest-scoring live runner-up, the Official MCP Registry (8.19 for this job). Smithery stays the last runner-up as a directory only. No score changed. Official registry as source of truth.","runnersUpIds":["glama","composio","pulse-mcp","aws-agent-registry","google-agent-registry","smithery"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"dx + capability. Official registry as source of truth (Depot pick since 2026-10-24; Smithery's connect/run services end 2026-10-31).","tags":["mcp"]},{"id":"progressive-agent-rollout","slug":"progressive-agent-rollout","name":"Progressive rollout of agent prompts, models, and tools","shortName":"Progressive rollout of agent prompts, models, and tools","problemStatement":"Ship a new prompt, model, or tool to 1% of agent traffic, eval online, and auto-rollback without a deploy.","whenNotTopPick":"Want cheaper flags+experiments in one product and can tolerate Amplitude-era Statsig packaging — Statsig; only need offline evals not traffic splitting — Braintrust/Promptfoo.","categoryIds":["cat-observability"],"depotPickProviderId":"launchdarkly","depotPickRationale":"reliability of flag delivery + agent-specific DX. LaunchDarkly AgentControl is the purpose-built SKU; Statsig wins pricePerf. Distinct from evals (CI quality) — this is production traffic control.","runnersUpIds":["statsig","braintrust","langfuse","growthbook","unleash","flagsmith","posthog"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"reliability of flag delivery + agent-specific DX. LaunchDarkly AgentControl is the purpose-built SKU; Statsig wins pricePerf. Distinct from evals (CI quality) — this is production traffic control.","tags":["progressive"]},{"id":"sql-warehouse-agent-context","slug":"sql-warehouse-agent-context","name":"SQL warehouse as agent context","shortName":"SQL warehouse as agent context","problemStatement":"Let agents query production analytics (SQL, Iceberg/DuckLake, shared semantic Guides) rather than dumping a warehouse into a vector DB.","whenNotTopPick":"Snowflake is already the system of record — Cortex; Databricks lakehouse standard — Mosaic; local/embedded only — DuckDB OSS.","categoryIds":["cat-vector-memory","cat-cpu-compute"],"depotPickProviderId":"motherduck","depotPickRationale":"DX for programmatic SQL + pricePerf of bursty agent queries. MotherDuck Lite/Pulse is the agent-native default; hyperscaler warehouses when the data already lives there.","runnersUpIds":["snowflake-cortex","databricks-mosaic","duckdb","clickhouse-cloud"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"DX for programmatic SQL + pricePerf of bursty agent queries. MotherDuck Lite/Pulse is the agent-native default; hyperscaler warehouses when the data already lives there.","tags":["sql"]},{"id":"prompt-injection-firewall","slug":"prompt-injection-firewall","name":"Prompt-injection / jailbreak firewall","shortName":"Prompt-injection / jailbreak firewall","problemStatement":"Screen untrusted user and retrieved content for instruction-override before it reaches the model — a dedicated firewall, not a content taxonomy.","whenNotTopPick":"Need programmable multi-step rails / topical policies — NeMo Guardrails; only need free hate/sexual/violence flags — OpenAI Moderation; already standardized on Palo Alto — Prisma AIRS; Cisco estate — cisco-ai-defense.","categoryIds":["cat-observability","cat-inference-apis"],"depotPickProviderId":"lakera","depotPickRationale":"capability on injection/jailbreak detection + dx of a single API. Distinct from content-moderation (taxonomy) and llm-input-guardrails (programmable policy engine).","runnersUpIds":["nvidia-nemo-guardrails","openai-moderation","prisma-airs","guardrails-ai","hiddenlayer"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"capability on injection/jailbreak detection + dx of a single API. Distinct from content-moderation (taxonomy) and llm-input-guardrails (programmable policy engine).","tags":["prompt"]},{"id":"agent-cloud-finops","slug":"agent-cloud-finops","name":"Unified cloud + GPU + LLM FinOps","shortName":"Unified cloud + GPU + LLM FinOps","problemStatement":"Allocate OpenAI/Anthropic/Modal/GPU spend per agent, customer, or feature next to the AWS bill — not just token traces in a gateway.","whenNotTopPick":"Only need request-path token tracing — Helicone; finance wants AI unit-economics and will buy a sales-led platform — CloudZero.","categoryIds":["cat-observability"],"depotPickProviderId":"vantage","depotPickRationale":"pricePerf of the FinOps tool itself + coverage of LLM/GPU connectors. Distinct from cost-monitoring (request-path token traces; pick Langfuse since 2026-10-21).","runnersUpIds":["helicone","cloudzero","langfuse"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf of the FinOps tool itself + coverage of LLM/GPU connectors. Distinct from cost-monitoring (request-path token traces; pick Langfuse since 2026-10-21).","tags":["agent"]},{"id":"serverless-agent-queue","slug":"serverless-agent-queue","name":"Serverless agent event backbone (queue/bus)","shortName":"Serverless agent event backbone (queue/bus)","problemStatement":"Need durable delayed HTTP callbacks, fan-out, or a real message fabric for agents — not a workflow runtime.","whenNotTopPick":"Need durable multi-step state/retries as code — Inngest/Temporal; need NATS subjects/leaf nodes — Synadia; need Kafka protocol/replay — WarpStream (or Confluent if that's already standard).","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"upstash-qstash","depotPickRationale":"DX + pricePerf for a bus that agents can fire-and-forget. Distinct from durable-agent-workflows (Temporal/Inngest steps) and event-driven-agent-jobs (Inngest DX default).","runnersUpIds":["inngest","synadia-cloud","warpstream","aws-sqs","aws-eventbridge","azure-service-bus"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"DX + pricePerf for a bus that agents can fire-and-forget. Distinct from durable-agent-workflows (Temporal/Inngest steps) and event-driven-agent-jobs (Inngest DX default).","tags":["serverless"]},{"id":"a2a-agent-discovery","slug":"a2a-agent-discovery","name":"Discover and govern A2A/multi-agent registries","shortName":"Discover and govern A2A/multi-agent registries","problemStatement":"Need a registry to publish/find Agent Cards (A2A) and related MCP/skills across teams or the public internet — not just an MCP directory.","whenNotTopPick":"Not on AWS — Google Agent Registry if GCP; public/OSS multi-vendor — OpenAgora or a2a-registry.dev; MCP-only directories — Smithery/Glama/PulseMCP/official registry.","categoryIds":["cat-search-scrape","cat-sandboxes-browsers"],"depotPickProviderId":"aws-agent-registry","depotPickRationale":"Prefer governed enterprise catalogs when IAM/approvals matter; open registries when protocol interop and zero lock-in matter. Distinct from mcp-discovery.","runnersUpIds":["google-agent-registry","openagora","a2a-registry","pulse-mcp"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Prefer governed enterprise catalogs when IAM/approvals matter; open registries when protocol interop and zero lock-in matter. Distinct from mcp-discovery.","tags":["a2a"]},{"id":"oss-feature-flags","slug":"oss-feature-flags","name":"Self-host / OSS feature flags for agent rollouts","shortName":"Self-host / OSS feature flags for agent rollouts","problemStatement":"Need progressive delivery and kill-switches for agent tools/prompts without LaunchDarkly/Statsig seat economics — preferably OSS or cheap seats.","whenNotTopPick":"Enterprise change-mgmt/approvals at scale — LaunchDarkly; flags-first audited self-host with Edge — Unleash; OpenFeature-shaped simpler cloud SKU — Flagsmith; already on Statsig/Amplitude — stay.","categoryIds":["cat-observability"],"depotPickProviderId":"growthbook","depotPickRationale":"PricePerf + privacy for self-host. Distinct from progressive-agent-rollout (commercial default LD).","runnersUpIds":["unleash","flagsmith","launchdarkly","statsig","posthog"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"PricePerf + privacy for self-host. Distinct from progressive-agent-rollout (commercial default LD).","tags":["oss"]},{"id":"hyperscaler-agent-queue","slug":"hyperscaler-agent-queue","name":"Hyperscaler queue/bus for agents already on AWS","shortName":"Hyperscaler queue/bus for agents already on AWS","problemStatement":"Agent callbacks, fan-out, and cron need to live inside AWS IAM/VPC — not a third-party HTTP queue.","whenNotTopPick":"Want HTTP-first serverless without AWS — QStash; durable multi-step code — Inngest/Temporal; pure competing consumers — SQS.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"aws-eventbridge","depotPickRationale":"Reliability + maturity inside AWS. Distinct from serverless-agent-queue (QStash default).","runnersUpIds":["aws-sqs","upstash-qstash","inngest","azure-service-bus","gcp-pubsub"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Reliability + maturity inside AWS. Distinct from serverless-agent-queue (QStash default).","tags":["hyperscaler"]},{"id":"llm-ai-gateway","slug":"llm-ai-gateway","name":"Production LLM AI gateway","shortName":"Production LLM AI gateway","problemStatement":"Need a unified gateway in front of many model providers with fallbacks, virtual keys, budgets, guardrails, and request logs — not just a model marketplace.","whenNotTopPick":"You only need a multi-model catalog/router (OpenRouter), pure cheap request logs (Helicone), or are standardized on Palo Alto Prisma AIRS enterprise gateway.","categoryIds":["cat-observability","cat-inference-apis"],"depotPickProviderId":"portkey","depotPickRationale":"DX + reliability + capability for fallbacks/keys/guardrails. OpenRouter wins catalog breadth; Portkey wins production gateway controls.","runnersUpIds":["vercel-ai-gateway","cloudflare-ai-gateway","openrouter","helicone","cloudflare-workers-ai"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"DX + reliability + capability for fallbacks/keys/guardrails. OpenRouter wins catalog breadth; Portkey wins production gateway controls.","tags":["llm"]},{"id":"multilingual-stt","slug":"multilingual-stt","name":"Multilingual / code-switching speech-to-text","shortName":"Multilingual / code-switching speech-to-text","problemStatement":"Agents must transcribe 20–50+ languages or mid-utterance language switches with production accuracy — not English-only streaming.","whenNotTopPick":"English-first agent economics dominate (Deepgram) or you need OpenAI Realtime single-vendor stack.","categoryIds":["cat-media-gen","cat-inference-apis"],"depotPickProviderId":"speechmatics","depotPickRationale":"Capability on language coverage + code-switch + privacy/on-prem. PricePerf secondary to accuracy across locales.","runnersUpIds":["gladia","deepgram","assemblyai"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability on language coverage + code-switch + privacy/on-prem. PricePerf secondary to accuracy across locales.","tags":["multilingual"]},{"id":"scrape-actor-platform","slug":"scrape-actor-platform","name":"Scraper Actor / marketplace platform","shortName":"Scraper Actor / marketplace platform","problemStatement":"Need a marketplace of reusable scrapers (Actors), scheduling, proxies, and agent-payable runs — not a single crawl-to-markdown API.","whenNotTopPick":"You only need LLM-ready crawl/extract (Firecrawl) or a simple anti-bot request API (ZenRows/Scrapfly).","categoryIds":["cat-search-scrape","cat-sandboxes-browsers"],"depotPickProviderId":"apify","depotPickRationale":"Capability of store + proxy + agent payment DX. Firecrawl remains llm-ready-extract pick.","runnersUpIds":["bright-data","scrapfly","zenrows"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability of store + proxy + agent payment DX. Firecrawl remains llm-ready-extract pick.","tags":["scrape"]},{"id":"cross-session-user-memory","slug":"cross-session-user-memory","name":"Cross-session / portable user memory","shortName":"Cross-session / portable user memory","problemStatement":"Need durable user-profile or portable memory that follows the user across sessions and assistants (often MCP), not just a vector DB or fact store.","whenNotTopPick":"You want the simplest hosted fact API (Mem0), temporal knowledge graphs (Zep), or a full MemGPT-style runtime (Letta).","categoryIds":["cat-vector-memory"],"depotPickProviderId":"memobase","depotPickRationale":"Capability of profile/portable memory + DX (MCP) + pricePerf. Distinct from rag-memory (pgvector) and stateful-agent-runtime (Letta).","runnersUpIds":["mem0","supermemory","zep","cognee"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability of profile/portable memory + DX (MCP) + pricePerf. Distinct from rag-memory (pgvector) and stateful-agent-runtime (Letta).","tags":["cross"]},{"id":"document-parse","slug":"document-parse","name":"Parse documents into LLM-ready chunks","shortName":"Parse documents into LLM-ready chunks","problemStatement":"Turn PDFs, Office, scans, and messy forms into markdown/JSON an agent can retrieve or extract from — not a web crawler.","whenNotTopPick":"Need connector-heavy ETL / VPC pipelines or OSS self-host partitioner — Unstructured; agentic forms/tables still fail LlamaParse bake-off — Reducto; you only have URLs not files — Firecrawl/Jina Reader.","categoryIds":["cat-search-scrape","cat-train-finetune"],"depotPickProviderId":"llama-parse","depotPickRationale":"Capability on tables/forms + dx + pricePerf of the cheap tier. Distinct from llm-ready-extract (web) and rag-memory (store).","runnersUpIds":["unstructured","reducto","jina"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability on tables/forms + dx + pricePerf of the cheap tier. Distinct from llm-ready-extract (web) and rag-memory (store).","tags":["document"]},{"id":"rag-rerank","slug":"rag-rerank","name":"Rerank retrieved chunks for RAG","shortName":"Rerank retrieved chunks for RAG","problemStatement":"After vector search, need a dedicated reranker so the agent sees the right 5–20 chunks — not just 'better embeddings'.","whenNotTopPick":"Already on Cohere embed+rerank in one bill — stay; want EU all-in-one embed/rerank/Reader — Jina; tiny corpus where embeddings alone suffice — skip rerank.","categoryIds":["cat-inference-apis","cat-vector-memory"],"depotPickProviderId":"voyage","depotPickRationale":"Capability of rerank quality + pricePerf (lite SKUs, free tokens). Distinct from embeddings (first-stage vectors).","runnersUpIds":["cohere","jina"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability of rerank quality + pricePerf (lite SKUs, free tokens). Distinct from embeddings (first-stage vectors).","tags":["rag"]},{"id":"mobile-computer-use","slug":"mobile-computer-use","name":"Real-device mobile computer-use","shortName":"Real-device mobile computer-use","problemStatement":"Agent must tap/swipe a real iOS or Android device (Appium/UIAutomator), not a desktop VM or headless Chrome.","whenNotTopPick":"Only need Android VMs / multi-OS desktop fleets — Cua; already on AWS and want $0.17/min metered or $250 slots — Device Farm; browser-only CUA — Anchor/Hyperbrowser; Linux desktop — E2B.","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"browserstack","depotPickRationale":"Capability of real-device catalog + reliability + dx (Appium). Distinct from desktop-computer-use and computer-use-browser.","runnersUpIds":["aws-device-farm","cua","sauce-labs","testmu-ai","genymotion-cloud"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability of real-device catalog + reliability + dx (Appium). Distinct from desktop-computer-use and computer-use-browser.","tags":["mobile"]},{"id":"realtime-agent-channel","slug":"realtime-agent-channel","name":"Realtime agent↔user channel","shortName":"Realtime agent↔user channel","problemStatement":"Stream agent tokens, tool progress, and presence to clients over WebSocket/SSE with pub/sub semantics — not email/SMS and not a backend work queue.","whenNotTopPick":"Need full chat UX (channels, moderation, reactions) — Stream; only server-side job fanout — QStash/SQS/EventBridge; Cloudflare Durable Objects already hold your edge state.","categoryIds":["cat-comms","cat-cpu-compute"],"depotPickProviderId":"ably","depotPickRationale":"Reliability + dx for realtime fanout. Distinct from notification-orchestration and serverless-agent-queue.","runnersUpIds":["getstream","cloudflare-workers","supabase-pgvector"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Reliability + dx for realtime fanout. Distinct from notification-orchestration and serverless-agent-queue.","tags":["realtime"]},{"id":"edge-sqlite-agent-db","slug":"edge-sqlite-agent-db","name":"Edge / per-tenant SQLite for agents","shortName":"Edge / per-tenant SQLite for agents","problemStatement":"Need many lightweight per-user or per-agent SQL databases at the edge (libSQL/SQLite), not a single Postgres or Redis cache.","whenNotTopPick":"Want shared serverless Postgres branching — Neon; Auth+Storage suite — Supabase; sub-ms shared KV — Redis.","categoryIds":["cat-vector-memory","cat-object-storage"],"depotPickProviderId":"turso","depotPickRationale":"pricePerf + dx for numerous small DBs. Distinct from serverless-db (Postgres) and agent-memory-kv (Redis).","runnersUpIds":["cloudflare-workers","neon","supabase-pgvector"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"pricePerf + dx for numerous small DBs. Distinct from serverless-db (Postgres) and agent-memory-kv (Redis).","tags":["edge"]},{"id":"ai-merchant-of-record","slug":"ai-merchant-of-record","name":"Sell AI/agent products as Merchant of Record","shortName":"Sell AI/agent products as Merchant of Record","problemStatement":"Need to charge for tokens, agent runs, seats, and credits with global tax/VAT handled — without building Stripe Tax + meters + invoicing yourself.","whenNotTopPick":"You already have finance/ops owning Stripe Tax + OpenMeter/Metronome, or you must stay pure payment-processor (not MoR).","categoryIds":["cat-auth-payments"],"depotPickProviderId":"polar","depotPickRationale":"DX + pricePerf + maturity. Polar wins public MoR+usage ladder for AI indie/agent products; Stripe remains payments default.","runnersUpIds":["stripe","openmeter","metronome","autumn","lemon-squeezy","paddle"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"DX + pricePerf + maturity. Polar wins public MoR+usage ladder for AI indie/agent products; Stripe remains payments default.","tags":["sell"]},{"id":"serverless-edge-redis","slug":"serverless-edge-redis","name":"Serverless HTTP Redis from edge/agents","shortName":"Serverless HTTP Redis from edge/agents","problemStatement":"Need Redis-compatible KV/cache/rate-limits from Workers/Vercel/Lambda without TCP connection pools or always-on memory bills.","whenNotTopPick":"Need fat always-on Redis modules / Redis Enterprise features — then Redis Cloud; pure edge state in CF — Durable Objects.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"upstash-redis","depotPickRationale":"DX + pricePerf + reliability for bursty serverless agents. Upstash HTTP Redis is the default.","runnersUpIds":["redis","cloudflare-workers","neon"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"DX + pricePerf + reliability for bursty serverless agents. Upstash HTTP Redis is the default.","tags":["serverless"]},{"id":"knowledge-graph-extract","slug":"knowledge-graph-extract","name":"Structured web → knowledge graph extract","shortName":"Structured web → knowledge graph extract","problemStatement":"Need schema'd entities/products/articles from the public web (not just markdown chunks) for agent grounding.","whenNotTopPick":"You only need LLM-ready markdown/crawl — Firecrawl/Jina; semantic browser queries — AgentQL.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"diffbot","depotPickRationale":"Capability for structured Extract+KG + pricePerf. Diffbot leads KG; Firecrawl remains llm-ready-extract pick.","runnersUpIds":["firecrawl","jina","unstructured","agentql"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability for structured Extract+KG + pricePerf. Diffbot leads KG; Firecrawl remains llm-ready-extract pick.","tags":["knowledge"]},{"id":"outbound-webhooks","slug":"outbound-webhooks","name":"Reliable outbound webhooks for agent products","shortName":"Reliable outbound webhooks for agent products","problemStatement":"Need to deliver signed, retried webhooks to customers (agent run finished, invoice, tool events) without building a queue+portal yourself.","whenNotTopPick":"You mainly need to *consume* flaky inbound provider webhooks — Hookdeck. Pure internal async — QStash/Inngest/EventBridge.","categoryIds":["cat-comms","cat-cpu-compute"],"depotPickProviderId":"svix","depotPickRationale":"Reliability + DX + maturity for customer-facing webhook delivery. Svix wins; Hookdeck is the inbound gateway peer.","runnersUpIds":["hookdeck","inngest","upstash-qstash","aws-eventbridge"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Reliability + DX + maturity for customer-facing webhook delivery. Svix wins; Hookdeck is the inbound gateway peer.","tags":["outbound"]},{"id":"saas-oauth-integrations","slug":"saas-oauth-integrations","name":"Embed SaaS OAuth + API sync for agents","shortName":"Embed SaaS OAuth + API sync for agents","problemStatement":"Need customers to connect HubSpot/Jira/Notion/etc. with durable OAuth tokens and syncs — not just one-shot MCP tool calls.","whenNotTopPick":"You need the widest MCP tool router for agents — Composio. Enterprise SSO only — WorkOS. Human workflow glue — Pipedream/Zapier.","categoryIds":["cat-sandboxes-browsers","cat-auth-payments"],"depotPickProviderId":"nango","depotPickRationale":"Capability for OAuth/sync embedding + self-host option. Nango wins productized integrations; Composio remains agent-tool-router pick.","runnersUpIds":["composio","workos"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability for OAuth/sync embedding + self-host option. Nango wins productized integrations; Composio remains agent-tool-router pick.","tags":["saas"]},{"id":"agent-analytics-warehouse","slug":"agent-analytics-warehouse","name":"High-volume agent event/log analytics warehouse","shortName":"High-volume agent event/log analytics warehouse","problemStatement":"Need OLAP SQL over millions of agent traces, tool calls, and product events — not just an app Postgres or DuckDB laptop.","whenNotTopPick":"Lightweight embedded analytics — MotherDuck/DuckDB. Already on Snowflake/Databricks — stay. Log search UI — Elastic/Datadog.","categoryIds":["cat-vector-memory","cat-observability","cat-cpu-compute"],"depotPickProviderId":"clickhouse-cloud","depotPickRationale":"Capability + pricePerf for analytical volume. ClickHouse Cloud wins agent event OLAP; MotherDuck remains sql-warehouse-agent-context pick.","runnersUpIds":["motherduck","snowflake-cortex","databricks-mosaic","elastic-serverless"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"Capability + pricePerf for analytical volume. ClickHouse Cloud wins agent event OLAP; MotherDuck remains sql-warehouse-agent-context pick.","tags":["agent"]},{"id":"agent-error-observability","slug":"agent-error-observability","name":"Application error/trace monitoring for agent backends","shortName":"Application error/trace monitoring for agent backends","problemStatement":"Need exception tracking, performance spans, and alerts for the services hosting agents — separate from LLM request tracing.","whenNotTopPick":"LLM-only tracing/evals — Langfuse/Braintrust/Helicone. Full enterprise APM already Datadog — stay.","categoryIds":["cat-observability"],"depotPickProviderId":"sentry","depotPickRationale":"DX + maturity for app errors. Sentry wins; keep LLM tracers separate.","runnersUpIds":["honeycomb","posthog"],"weights":{"capabilityFit":0.25,"pricePerformance":0.18,"latencyReliability":0.15,"developerDx":0.15,"privacyPosture":0.1,"lockInRisk":0.07,"maturity":0.1},"methodologyNotes":"DX + maturity for app errors. Sentry wins; keep LLM tracers separate.","tags":["agent"]},{"id":"confidential-tee-inference","slug":"confidential-tee-inference","name":"Confidential / TEE-backed LLM inference","shortName":"TEE inference","problemStatement":"Run inference where the host cannot read prompts/weights in cleartext — confidential VMs or TEE-backed serverless for regulated or high-trust agents.","whenNotTopPick":"Need hyperscaler confidential-compute contracts with BAA/DPA already negotiated (Bedrock/Azure), or a dedicated Friendli/GPU TEE SKU your security team already approved.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"chutes","depotPickRationale":"Capability = real TEE/confidential posture + model coverage. Privacy weighted high. Chutes publishes a TEE matrix agents can pin; hyperscalers win when enterprise paperwork dominates. Confidence: medium — re-curl chutes.ai/pricing TEE rows before production.","runnersUpIds":["friendli","amazon-bedrock","azure-openai","fireworks"],"weights":{"capabilityFit":0.2,"pricePerformance":0.1,"latencyReliability":0.12,"developerDx":0.1,"privacyPosture":0.3,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Capability = real TEE/confidential posture + model coverage. Privacy weighted high. Chutes publishes a TEE matrix agents can pin; hyperscalers win when enterprise paperwork dominates. Confidence: medium — re-curl chutes.ai/pricing TEE rows before production.","tags":["tee","confidential","privacy","inference"]},{"id":"batch-cheap-embeddings","slug":"batch-cheap-embeddings","name":"High-volume cheap batch embeddings","shortName":"Batch embeddings","problemStatement":"Embed millions of chunks overnight where per-token quality matters less than $/1M and throughput; async/batch APIs preferred.","whenNotTopPick":"Need multilingual/EU-hosted embedders first (Jina), Cohere embed+rerank bundle, or Gemini/OpenAI already in the same bill.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"voyage","depotPickRationale":"pricePerf + reliability of batch jobs. Distinct from `embeddings` (quality-first RAG). Confidence: high on Voyage/Jina public cards; verify batch SKU names.","runnersUpIds":["jina","cohere","openai","google-ai"],"weights":{"capabilityFit":0.18,"pricePerformance":0.32,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"pricePerf + reliability of batch jobs. Distinct from `embeddings` (quality-first RAG). Confidence: high on Voyage/Jina public cards; verify batch SKU names.","tags":["embeddings","batch","rag","cost"]},{"id":"continuous-eval-ci","slug":"continuous-eval-ci","name":"Continuous eval gates in CI/CD","shortName":"Eval CI gates","problemStatement":"Block merges when agent prompts/tools regress — hosted eval suites, datasets, and pass/fail gates wired to GitHub/GitLab CI.","whenNotTopPick":"Want OSS-first local gates (Promptfoo), DeepEval/Confident already standardized, or LangSmith is the org tracing system of record.","categoryIds":["cat-observability"],"depotPickProviderId":"braintrust","depotPickRationale":"Distinct from `evals` (general harness): CI gate DX + dataset versioning + flake control. Confidence: high for Braintrust/Promptfoo CI docs.","runnersUpIds":["promptfoo","confident-ai","langsmith","patronus"],"weights":{"capabilityFit":0.22,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.28,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.08},"methodologyNotes":"Distinct from `evals` (general harness): CI gate DX + dataset versioning + flake control. Confidence: high for Braintrust/Promptfoo CI docs.","tags":["evals","ci","regression","agents"]},{"id":"prompt-caching-optimization","slug":"prompt-caching-optimization","name":"Prompt-caching cost optimization","shortName":"Prompt caching","problemStatement":"Cut spend on long system prompts, tool schemas, and RAG prefixes via provider prompt caching / context caching with predictable hit rates.","whenNotTopPick":"Workloads are short-prefix high-variance (cache thrash), or gateway-level cache across vendors (Portkey/CF AI Gateway) matters more than first-party cache.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"anthropic","depotPickRationale":"pricePerf + DX of cache controls + hit-rate observability. Anthropic prompt caching is the clearest agent-facing story; OpenAI/Gemini close. Confidence: high.","runnersUpIds":["openai","google-ai","portkey","amazon-bedrock"],"weights":{"capabilityFit":0.18,"pricePerformance":0.32,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"pricePerf + DX of cache controls + hit-rate observability. Anthropic prompt caching is the clearest agent-facing story; OpenAI/Gemini close. Confidence: high.","tags":["caching","cost","prompts","inference"]},{"id":"agent-pstn-telephony","slug":"agent-pstn-telephony","name":"PSTN / phone-number telephony for agents","shortName":"Agent telephony","problemStatement":"Provision numbers, place/receive PSTN calls, and bridge SIP/WebRTC so voice agents can dial real phones — not just in-browser audio.","whenNotTopPick":"Want an all-in-one voice-agent stack (Vapi/Retell) instead of raw telephony, or Telnyx pricing/regions fit better.","categoryIds":["cat-comms"],"depotPickProviderId":"twilio","depotPickRationale":"Distinct from `voice-agent-platform` (orchestration+STT+TTS). Here capability = numbers, regulatory, SIP, global reach. Confidence: high.","runnersUpIds":["telnyx","vapi","retell","plivo"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.28,"developerDx":0.14,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from `voice-agent-platform` (orchestration+STT+TTS). Here capability = numbers, regulatory, SIP, global reach. Confidence: high.","tags":["telephony","pstn","voice","cpaas"]},{"id":"calendar-scheduling-tools","slug":"calendar-scheduling-tools","name":"Calendar & scheduling tools for agents","shortName":"Calendar tools","problemStatement":"Agents must read/write Google/Microsoft calendars, propose slots, and book meetings via OAuth toolkits without building calendar OAuth from scratch.","whenNotTopPick":"You only need generic SaaS sync (Nango), or enterprise directory/SSO is the hard part (WorkOS) rather than calendar verbs.","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"composio","depotPickRationale":"No first-party Calendar SaaS in Depot — pick is tool-router based. Confidence: medium; verify Composio calendar actions + scopes before pilot.","runnersUpIds":["nango","workos","trigger-dev"],"weights":{"capabilityFit":0.22,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.28,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.08},"methodologyNotes":"No first-party Calendar SaaS in Depot — pick is tool-router based. Confidence: medium; verify Composio calendar actions + scopes before pilot.","tags":["calendar","scheduling","oauth","tools","mcp"]},{"id":"layout-aware-ocr","slug":"layout-aware-ocr","name":"Layout-aware OCR & form extraction","shortName":"Layout OCR","problemStatement":"Extract tables, checkboxes, multi-column layouts, and key-value forms from PDFs/scans where naive chunking destroys structure.","whenNotTopPick":"Simple markdown chunking is enough (LlamaParse), or you need broad ETL connectors (Unstructured).","categoryIds":["cat-search-scrape"],"depotPickProviderId":"reducto","depotPickRationale":"Distinct from `document-parse` (general LLM-ready chunks). Weight capability on layout fidelity. Confidence: medium-high on Reducto agentic OCR positioning.","runnersUpIds":["llama-parse","unstructured","jina"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from `document-parse` (general LLM-ready chunks). Weight capability on layout fidelity. Confidence: medium-high on Reducto agentic OCR positioning.","tags":["ocr","pdf","forms","document-ai"]},{"id":"spreadsheet-sql-agents","slug":"spreadsheet-sql-agents","name":"Spreadsheet / CSV agents with SQL","shortName":"Sheet SQL agents","problemStatement":"Let agents query uploads, Google-sheet exports, and warehouse-like CSVs with SQL + LLM planning instead of brittle cell scraping.","whenNotTopPick":"Need Sheets writeback via OAuth tools (Composio), embedded local DuckDB only, or full lakehouse (Databricks).","categoryIds":["cat-vector-memory"],"depotPickProviderId":"motherduck","depotPickRationale":"MotherDuck = hosted DuckDB with agent/SQL DX. Confidence: medium — confirm MCP/SQL agent features on current pricing.","runnersUpIds":["duckdb","composio","neon","databricks-mosaic"],"weights":{"capabilityFit":0.22,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.28,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.08},"methodologyNotes":"MotherDuck = hosted DuckDB with agent/SQL DX. Confidence: medium — confirm MCP/SQL agent features on current pricing.","tags":["spreadsheet","sql","duckdb","analytics"]},{"id":"diff-aware-pr-review","slug":"diff-aware-pr-review","name":"Diff-aware pull-request review agents","shortName":"PR review agents","problemStatement":"Review multi-file diffs with high tool-use reliability, low false-positive nits, and safe suggestions for CI review bots.","whenNotTopPick":"Bulk low-cost review of huge monorepos (DeepSeek/open hosts) with human confirm, or procurement locks you to Azure OpenAI.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"anthropic","depotPickRationale":"Related to `coding-agent` but optimized for review posture (precision, rationale, patch safety). Confidence: high on model quality; product wrappers vary.","runnersUpIds":["openai","google-ai","deepseek","mistral"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Related to `coding-agent` but optimized for review posture (precision, rationale, patch safety). Confidence: high on model quality; product wrappers vary.","tags":["code-review","pr","coding","agents"]},{"id":"secret-rotation-automation","slug":"secret-rotation-automation","name":"Automated secret rotation for agents","shortName":"Secret rotation","problemStatement":"Rotate API keys, DB creds, and workload identities on a schedule with zero-downtime dual-publish so long-running agents do not die mid-job.","whenNotTopPick":"Need self-host/OSS (Infisical), AWS-native rotation Lambdas, or Vault already mandated.","categoryIds":["cat-auth-payments"],"depotPickProviderId":"doppler","depotPickRationale":"Distinct from `agent-secrets` (storage/injection). Here maturity of rotation workflows + DX. Confidence: high.","runnersUpIds":["infisical","aws-secrets-manager","hashicorp-vault","onepassword-secrets"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.28,"developerDx":0.14,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from `agent-secrets` (storage/injection). Here maturity of rotation workflows + DX. Confidence: high.","tags":["secrets","rotation","security","agents"]},{"id":"llm-audit-log-warehouse","slug":"llm-audit-log-warehouse","name":"LLM audit-log / compliance warehouse","shortName":"LLM audit warehouse","problemStatement":"Retain prompt/tool/decision logs for SOC2/ISO auditors in a queryable warehouse with retention, access control, and export — not just a tracing UI.","whenNotTopPick":"Product-analytics-style tracing UX first (Langfuse/Honeycomb), or Elastic already is the SIEM.","categoryIds":["cat-observability"],"depotPickProviderId":"clickhouse-cloud","depotPickRationale":"Observability category; privacy + maturity. ClickHouse wins for high-cardinality LLM logs at cost. Confidence: medium on turnkey LLM schemas.","runnersUpIds":["honeycomb","langfuse","sentry","elastic-serverless"],"weights":{"capabilityFit":0.2,"pricePerformance":0.1,"latencyReliability":0.12,"developerDx":0.1,"privacyPosture":0.3,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Observability category; privacy + maturity. ClickHouse wins for high-cardinality LLM logs at cost. Confidence: medium on turnkey LLM schemas.","tags":["compliance","audit","siem","logs"]},{"id":"synthetic-eval-data","slug":"synthetic-eval-data","name":"Synthetic evaluation & scenario data","shortName":"Synthetic eval data","problemStatement":"Generate diverse agent test scenarios, adversarial prompts, and labeled trajectories without waiting on production traffic.","whenNotTopPick":"Need open-model bulk generation (Fireworks), or Promptfoo red-team generators already cover the matrix.","categoryIds":["cat-train-finetune"],"depotPickProviderId":"openai","depotPickRationale":"No Scale/Labelbox vendor in catalog — LLM-synthetic path only. Confidence: medium; human review still required for high-stakes evals.","runnersUpIds":["anthropic","google-ai","fireworks","promptfoo"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"No Scale/Labelbox vendor in catalog — LLM-synthetic path only. Confidence: medium; human review still required for high-stakes evals.","tags":["synthetic-data","evals","datasets"]},{"id":"lora-serving-distillation","slug":"lora-serving-distillation","name":"LoRA serving & distillation hosting","shortName":"LoRA / distill host","problemStatement":"Train adapters or distill smaller students, then serve many LoRAs efficiently without one-GPU-per-customer economics.","whenNotTopPick":"DIY Unsloth + your GPU cloud, Lightning Studios workflow, or Together/Fireworks finetune+serve already contracted.","categoryIds":["cat-train-finetune"],"depotPickProviderId":"predibase","depotPickRationale":"LoRAX-style multi-adapter serving is the differentiator. Confidence: medium — Predibase public token matrix uneven (see gaps).","runnersUpIds":["together","fireworks","unsloth","lightning-ai"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"LoRAX-style multi-adapter serving is the differentiator. Confidence: medium — Predibase public token matrix uneven (see gaps).","tags":["lora","finetune","distillation","serving"]},{"id":"agent-ephemeral-ci","slug":"agent-ephemeral-ci","name":"Ephemeral CI runners for agent jobs","shortName":"Agent CI runners","problemStatement":"Spin isolated CPU/GPU sandboxes per CI job to run agent integration tests, tool mocks, and eval suites that need real browsers or kernels.","whenNotTopPick":"Need browser-native sandboxes (E2B), Fly Machines already in mesh, or Upstash Box when GA fits.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"modal","depotPickRationale":"Serverless GPU/CPU job DX for CI. Confidence: high for Modal; Box still preview.","runnersUpIds":["e2b","fly-machines","daytona","upstash-box"],"weights":{"capabilityFit":0.22,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.28,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.08},"methodologyNotes":"Serverless GPU/CPU job DX for CI. Confidence: high for Modal; Box still preview.","tags":["ci","sandbox","serverless","agents"]},{"id":"visual-regression-qa","slug":"visual-regression-qa","name":"Screenshot & visual regression QA","shortName":"Visual QA","problemStatement":"Capture screenshots across browsers/devices and diff UI regressions for agent-driven release gates.","whenNotTopPick":"Agent browser automation is the product (Browserbase/Hyperbrowser), not device-farm coverage.","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"browserstack","depotPickRationale":"Device-farm breadth + SDKs. Confidence: high.","runnersUpIds":["browserbase","sauce-labs","testmu-ai","hyperbrowser"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.28,"developerDx":0.14,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Device-farm breadth + SDKs. Confidence: high.","tags":["qa","screenshots","visual-regression","devices"]},{"id":"form-filling-rpa","slug":"form-filling-rpa","name":"Form-filling & portal RPA agents","shortName":"Form-fill RPA","problemStatement":"Fill complex government/enterprise web forms with vision+DOM agents when no API exists — retries, selectors, and human-hand-off included.","whenNotTopPick":"Need stealth multi-session browsers without RPA opinionation (Anchor/Browserbase), or open-source self-host Browser Use.","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"skyvern","depotPickRationale":"Distinct from generic computer-use browsers: form/workflow success rate. Confidence: medium.","runnersUpIds":["anchor-browser","browser-use","hyperbrowser","browserbase"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from generic computer-use browsers: form/workflow success rate. Confidence: medium.","tags":["rpa","forms","computer-use","browser"]},{"id":"mcp-server-hosting","slug":"mcp-server-hosting","name":"Hosted MCP server runtime","shortName":"MCP hosting","problemStatement":"Deploy and scale MCP servers (tools/resources) with auth, logging, and HTTPS without babysitting containers — discovery alone is not enough.","whenNotTopPick":"Only need registry listing (mcp-registry/Glama), or Composio-managed tools replace custom MCP.","categoryIds":["cat-sandboxes-browsers"],"depotPickProviderId":"cloudflare-workers","depotPickRationale":"Distinct from `mcp-discovery`. Editorial replacement 2026-10-24: Smithery's own page (https://smithery.ai/future, fetched 2026-10-24) says 'Server URLs on run.tools and server.smithery.ai' stop working on October 31, 2026, so the former pick smithery (7.87 for this job) was replaced by the highest-scoring live runner-up, Cloudflare Workers (8.52 for this job), and removed from this job. No score changed. Confidence: medium — managed hosting SKUs move.","runnersUpIds":["glama","composio","fly-io"],"weights":{"capabilityFit":0.22,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.28,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.08},"methodologyNotes":"Distinct from `mcp-discovery`. Confidence: medium — managed hosting SKUs move; Smithery's hosted server URLs end 2026-10-31 (pick replaced 2026-10-24).","tags":["mcp","hosting","tools","runtime"]},{"id":"agent-m2m-identity","slug":"agent-m2m-identity","name":"Machine-to-machine agent identity","shortName":"Agent M2M identity","problemStatement":"Issue non-human identities, mTLS/JWT, and scoped tokens so agents authenticate to APIs without sharing user sessions.","whenNotTopPick":"Enterprise SSO/SCIM is the buyer (WorkOS/Auth0), or secrets injection is the only need (Infisical).","categoryIds":["cat-auth-payments"],"depotPickProviderId":"stytch","depotPickRationale":"Auth category; privacy + maturity. Not a full W3C DID network — pragmatic M2M. Confidence: medium-high.","runnersUpIds":["workos","auth0","clerk","infisical"],"weights":{"capabilityFit":0.2,"pricePerformance":0.1,"latencyReliability":0.12,"developerDx":0.1,"privacyPosture":0.3,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Auth category; privacy + maturity. Not a full W3C DID network — pragmatic M2M. Confidence: medium-high.","tags":["identity","m2m","auth","agents"]},{"id":"llm-rate-limit-shield","slug":"llm-rate-limit-shield","name":"LLM rate-limit & backoff shield","shortName":"Rate-limit shield","problemStatement":"Protect upstream model quotas with per-tenant limits, smart retries, fallbacks, and 429 shielding so one noisy agent cannot melt the budget.","whenNotTopPick":"Need per-virtual-key budgets and a full governance gateway (Portkey); Vercel already terminates AI traffic (Vercel AI Gateway); Helicone only if you accept maintenance-mode risk (since 2026-03-03, ownership unclear).","categoryIds":["cat-observability"],"depotPickProviderId":"cloudflare-ai-gateway","depotPickRationale":"Editorial replacement 2026-10-21: the former pick Helicone (8.12 for this job) has been in maintenance mode since 2026-03-03 (https://www.mintlify.com/blog/mintlify-acquires-helicone) and its site now says 'Helicone is no longer part of Mintlify. Details to come soon.' (https://www.helicone.ai/, re-checked 2026-10-21), so ownership and roadmap are unclear. Replaced by the highest-scoring live runner-up, Cloudflare AI Gateway (8.42 for this job; rate limiting is a free core feature, fixed or sliding windows, 429 on excess, per https://developers.cloudflare.com/ai-gateway/features/rate-limiting/ and /reference/pricing/; per user/project/team rate and budget limits with fallbacks via Dynamic Routing per /features/dynamic-routing/; all re-checked 2026-10-21; AI Gateway component 'operational' on cloudflarestatus.com 2026-10-21). Helicone stays a runner-up with the maintenance_mode flag. No score changed. Distinct from `llm-ai-gateway` (routing/guardrails). Focus rate limits + cost ceilings.","runnersUpIds":["portkey","vercel-ai-gateway","openrouter","helicone"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.28,"developerDx":0.14,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from `llm-ai-gateway` (routing/guardrails). Focus rate limits + cost ceilings. Confidence: high.","tags":["rate-limit","gateway","cost","reliability"]},{"id":"durable-idempotency","slug":"durable-idempotency","name":"Idempotency keys for agent side effects","shortName":"Agent idempotency","problemStatement":"Exactly-once-ish side effects (charges, tickets, emails) across retries using durable workflows and idempotency keys agents can pass end-to-end.","whenNotTopPick":"Temporal already standard, Inngest/TS DX preferred, or DBOS Postgres-centric apps.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"restate","depotPickRationale":"Distinct from generic `durable-agent-workflows`: emphasize idempotency primitives. Confidence: medium on Restate paid Cloud $/action.","runnersUpIds":["temporal-cloud","inngest","hatchet","dbos"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.28,"developerDx":0.14,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from generic `durable-agent-workflows`: emphasize idempotency primitives. Confidence: medium on Restate paid Cloud $/action.","tags":["idempotency","workflows","exactly-once","agents"]},{"id":"agent-dlq-retry","slug":"agent-dlq-retry","name":"Dead-letter queues for agent tasks","shortName":"Agent DLQ","problemStatement":"Park failed agent runs with full payload, replay, and poison-pill isolation so human operators can unblock queues without data loss.","whenNotTopPick":"AWS/Azure already mandated, or NATS/JetStream is the mesh (Synadia).","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"upstash-qstash","depotPickRationale":"Serverless HTTP DLQ DX for agents. Confidence: high for QStash; verify DLQ/replay knobs.","runnersUpIds":["aws-sqs","azure-service-bus","synadia-cloud","gcp-pubsub"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.28,"developerDx":0.14,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Serverless HTTP DLQ DX for agents. Confidence: high for QStash; verify DLQ/replay knobs.","tags":["dlq","queue","retry","agents"]},{"id":"hitl-approval-workflows","slug":"hitl-approval-workflows","name":"Human-in-the-loop approval workflows","shortName":"HITL approvals","problemStatement":"Pause agents for human approve/reject on risky tools (spend, delete, send), with timeouts, escalations, and audit trails.","whenNotTopPick":"Notification-only approval UX (Knock/Courier) without durable state, or Inngest wait-for-event fits better.","categoryIds":["cat-cpu-compute"],"depotPickProviderId":"temporal-cloud","depotPickRationale":"Workflow durability + signals. Confidence: high for Temporal patterns; productized HITL UIs vary.","runnersUpIds":["inngest","knock","courier","hatchet"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.28,"developerDx":0.14,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Workflow durability + signals. Confidence: high for Temporal patterns; productized HITL UIs vary.","tags":["hitl","approvals","workflows","safety"]},{"id":"llm-cost-anomaly-alerts","slug":"llm-cost-anomaly-alerts","name":"LLM spend anomaly alerts","shortName":"Cost anomaly alerts","problemStatement":"Detect sudden $/day spikes, runaway tool loops, and model-mix regressions — alert Slack/PagerDuty before the invoice lands.","whenNotTopPick":"LLM-only proxy metrics suffice (Helicone), or CloudZero is already the FinOps system of record.","categoryIds":["cat-observability"],"depotPickProviderId":"vantage","depotPickRationale":"Distinct from `agent-cloud-finops` (unified allocation) and `cost-monitoring` (dashboards). Focus anomaly detection. Confidence: medium on Vantage AI-specific detectors.","runnersUpIds":["helicone","cloudzero","langfuse","portkey"],"weights":{"capabilityFit":0.18,"pricePerformance":0.32,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"Distinct from `agent-cloud-finops` (unified allocation) and `cost-monitoring` (dashboards). Focus anomaly detection. Confidence: medium on Vantage AI-specific detectors.","tags":["finops","anomaly","alerts","cost"]},{"id":"low-carbon-gpu-inference","slug":"low-carbon-gpu-inference","name":"Low-carbon / region-aware GPU inference","shortName":"Low-carbon GPU","problemStatement":"Prefer cleaner energy regions or sustainable GPU clouds when agents can trade a few ms of latency for lower carbon intensity.","whenNotTopPick":"EU sovereign residency dominates (Scaleway/Gcore), or absolute cheapest spot (Salad) matters more than carbon claims.","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"crusoe","depotPickRationale":"Privacy/region + maturity of sustainability claims. Confidence: medium — verify Crusoe region energy disclosures for your ESG report.","runnersUpIds":["salad","verda","scaleway-generative","gcore"],"weights":{"capabilityFit":0.2,"pricePerformance":0.1,"latencyReliability":0.12,"developerDx":0.1,"privacyPosture":0.3,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Privacy/region + maturity of sustainability claims. Confidence: medium — verify Crusoe region energy disclosures for your ESG report.","tags":["carbon","sustainability","gpu","regions"]},{"id":"hipaa-baa-llm","slug":"hipaa-baa-llm","name":"HIPAA-friendly LLM with BAA path","shortName":"HIPAA LLM","problemStatement":"Process PHI-adjacent text with a provider willing to sign a BAA / equivalent and offer enterprise isolation controls.","whenNotTopPick":"Azure OpenAI already under Microsoft BAA, Vertex mandated, or memory-layer HIPAA (Supermemory) is the gap not the LLM.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"amazon-bedrock","depotPickRationale":"GUESS-adjacent: always confirm current BAA eligibility with counsel + vendor. Privacy weighted. Confidence: medium (legal, not just tech).","runnersUpIds":["azure-openai","vertex-ai","anthropic","supermemory"],"weights":{"capabilityFit":0.2,"pricePerformance":0.1,"latencyReliability":0.12,"developerDx":0.1,"privacyPosture":0.3,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"GUESS-adjacent: always confirm current BAA eligibility with counsel + vendor. Privacy weighted. Confidence: medium (legal, not just tech).","tags":["hipaa","baa","healthcare","compliance"]},{"id":"air-gapped-model-serve","slug":"air-gapped-model-serve","name":"Air-gapped / VPC model serving","shortName":"Air-gapped serve","problemStatement":"Serve optimized models inside VPC/air-gapped networks with no mandatory data egress to public LLM APIs.","whenNotTopPick":"Private Bedrock/Azure endpoints suffice without NIM packaging, or HF Inference Endpoints already in VPC.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"nvidia-nim","depotPickRationale":"Remove invalid runner if any. NIM for optimized self-host. Confidence: medium on packaging/licensing.","runnersUpIds":["huggingface-inference","amazon-bedrock","azure-openai"],"weights":{"capabilityFit":0.2,"pricePerformance":0.1,"latencyReliability":0.12,"developerDx":0.1,"privacyPosture":0.3,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Remove invalid runner if any. NIM for optimized self-host. Confidence: medium on packaging/licensing.","tags":["air-gapped","vpc","self-host","nim"]},{"id":"wasm-edge-small-models","slug":"wasm-edge-small-models","name":"Edge / WASM-adjacent small-model inference","shortName":"Edge small models","problemStatement":"Run small LLMs/embedders next to users on edge runtimes for privacy and TTFT — not full frontier chat.","whenNotTopPick":"Need LPU-class latency on larger open models (Groq), or Workers AI frontier SKUs require Workers Paid you will not buy.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"cloudflare-workers-ai","depotPickRationale":"Distinct from `edge-llm-inference` emphasis: small/on-device-adjacent catalog + WASM/isolates. Confidence: high for CF Workers AI neuron table freshness.","runnersUpIds":["groq","fireworks","deno-deploy","openrouter"],"weights":{"capabilityFit":0.18,"pricePerformance":0.32,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"Distinct from `edge-llm-inference` emphasis: small/on-device-adjacent catalog + WASM/isolates. Confidence: high for CF Workers AI neuron table freshness.","tags":["edge","wasm","small-models","privacy"]},{"id":"knowledge-graph-rag","slug":"knowledge-graph-rag","name":"Knowledge-graph / Graph-RAG memory","shortName":"Graph RAG","problemStatement":"Build entity-relation memory so agents answer multi-hop questions better than flat vector RAG alone.","whenNotTopPick":"Weaviate hybrid already enough, Zep temporal graphs fit sessions better, or Diffbot KG extract is the input stage only.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"cognee","depotPickRationale":"Confidence: medium — Cognee OSS/EU posture; confirm hosted SLA.","runnersUpIds":["weaviate","zep","diffbot"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Confidence: medium — Cognee OSS/EU posture; confirm hosted SLA.","tags":["graph-rag","knowledge-graph","memory"]},{"id":"late-interaction-retrieval","slug":"late-interaction-retrieval","name":"Late-interaction / ColBERT-style retrieval","shortName":"ColBERT retrieval","problemStatement":"Need multi-vector late interaction retrieval for high-recall RAG where single-vector ANN misses nuanced matches.","whenNotTopPick":"Standard embed+rerank (Voyage/Cohere) meets quality bar, or Elastic already stores the corpus.","categoryIds":["cat-inference-apis"],"depotPickProviderId":"jina","depotPickRationale":"Jina historically strong on late interaction / ColBERT-family. Confidence: medium — pin current model card.","runnersUpIds":["voyage","cohere","typesense-cloud"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Jina historically strong on late interaction / ColBERT-family. Confidence: medium — pin current model card.","tags":["colbert","retrieval","rag","multi-vector"]},{"id":"agent-trace-session-replay","slug":"agent-trace-session-replay","name":"Agent session trace replay","shortName":"Session replay","problemStatement":"Replay full agent sessions (LLM calls, tools, browser steps) for debugging incidents and training reviewers — like session replay for agents.","whenNotTopPick":"OpenTelemetry-first warehouse (Honeycomb), OSS Phoenix/Langfuse, or Braintrust eval-centric traces.","categoryIds":["cat-observability"],"depotPickProviderId":"langsmith","depotPickRationale":"Observability UX for multi-step agents. Confidence: high.","runnersUpIds":["braintrust","honeycomb","arize-phoenix","langfuse"],"weights":{"capabilityFit":0.22,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.28,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.08},"methodologyNotes":"Observability UX for multi-step agents. Confidence: high.","tags":["tracing","replay","debugging","agents"]},{"id":"runtime-llm-firewall","slug":"runtime-llm-firewall","name":"Runtime LLM firewall / AI WAF","shortName":"LLM firewall","problemStatement":"Inline policy enforcement on inputs/outputs (PII, jailbreaks, topic blocks) as a firewall layer in front of every model call.","whenNotTopPick":"Prompt-injection-only (Lakera), OSS validators (Guardrails AI), or gateway-native guards (Portkey).","categoryIds":["cat-observability"],"depotPickProviderId":"arthur-ai","depotPickRationale":"Distinct from `prompt-injection-firewall`: broader runtime WAF/policy. Confidence: medium — Arthur packaging/pricing.","runnersUpIds":["lakera","guardrails-ai","cisco-ai-defense","portkey"],"weights":{"capabilityFit":0.2,"pricePerformance":0.1,"latencyReliability":0.12,"developerDx":0.1,"privacyPosture":0.3,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from `prompt-injection-firewall`: broader runtime WAF/policy. Confidence: medium — Arthur packaging/pricing.","tags":["firewall","waf","safety","guardrails"]},{"id":"adversarial-red-team-evals","slug":"adversarial-red-team-evals","name":"Adversarial red-team evaluation suites","shortName":"Red-team evals","problemStatement":"Continuously attack prompts/agents with jailbreak corpora and safety suites before and after each release.","whenNotTopPick":"Patronus safety SKUs preferred, or Confident/DeepEval already embedded.","categoryIds":["cat-observability"],"depotPickProviderId":"promptfoo","depotPickRationale":"Distinct from `continuous-eval-ci` (quality gates) — safety adversarial focus. Confidence: high for Promptfoo red-team.","runnersUpIds":["patronus","confident-ai","galileo","lakera"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from `continuous-eval-ci` (quality gates) — safety adversarial focus. Confidence: high for Promptfoo red-team.","tags":["red-team","safety","evals","security"]},{"id":"ml-dataset-versioning","slug":"ml-dataset-versioning","name":"Dataset & experiment versioning for agents","shortName":"Dataset versioning","problemStatement":"Version eval sets, traces-as-data, and fine-tune corpora with lineage so agent improvements are reproducible.","whenNotTopPick":"LangSmith/Braintrust already store datasets next to traces, or Comet Opik is standard.","categoryIds":["cat-train-finetune"],"depotPickProviderId":"weights-biases","depotPickRationale":"MLOps maturity. Confidence: high for W&B; agent-specific dataset UX varies.","runnersUpIds":["langsmith","braintrust","comet-opik","langfuse"],"weights":{"capabilityFit":0.22,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.28,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.08},"methodologyNotes":"MLOps maturity. Confidence: high for W&B; agent-specific dataset UX varies.","tags":["datasets","versioning","mlops","lineage"]},{"id":"vector-object-hybrid-memory","slug":"vector-object-hybrid-memory","name":"Vector + object-storage hybrid memory","shortName":"Vector+blob memory","problemStatement":"Store embeddings cheaply atop object storage while keeping raw blobs (PDFs, images) co-located for agent re-read without double egress bills.","whenNotTopPick":"Managed Pinecone simplicity wins, or R2/Tigris alone if you bring your own ANN.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"turbopuffer","depotPickRationale":"Cost architecture pick. Confidence: high for Turbopuffer object-storage thesis.","runnersUpIds":["tigris","r2","pinecone","s3"],"weights":{"capabilityFit":0.18,"pricePerformance":0.32,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"Cost architecture pick. Confidence: high for Turbopuffer object-storage thesis.","tags":["vector","object-storage","memory","cost"]},{"id":"episodic-long-term-memory","slug":"episodic-long-term-memory","name":"Long-term episodic agent memory","shortName":"Episodic memory","problemStatement":"Remember user episodes across weeks with temporal validity (what was true when) — not just a static profile vector.","whenNotTopPick":"Mem0 simplicity, Letta stateful runtime, or HIPAA memory (Supermemory) is the constraint.","categoryIds":["cat-vector-memory"],"depotPickProviderId":"zep","depotPickRationale":"Distinct from `hosted-agent-memory` / `temporal-graph-memory` nuance: episodic timelines. Confidence: medium.","runnersUpIds":["letta","mem0","supermemory","memobase"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Distinct from `hosted-agent-memory` / `temporal-graph-memory` nuance: episodic timelines. Confidence: medium.","tags":["memory","episodic","temporal","agents"]},{"id":"inbound-webhook-gateway","slug":"inbound-webhook-gateway","name":"Inbound webhook gateway for agents","shortName":"Inbound webhooks","problemStatement":"Receive third-party webhooks reliably (verify, dedupe, queue, observe) and wake agents without exposing raw public endpoints.","whenNotTopPick":"Outbound-only (Svix), or EventBridge already central.","categoryIds":["cat-comms"],"depotPickProviderId":"hookdeck","depotPickRationale":"Pair to `outbound-webhooks`. Confidence: high.","runnersUpIds":["svix","inngest","aws-eventbridge","upstash-qstash"],"weights":{"capabilityFit":0.2,"pricePerformance":0.12,"latencyReliability":0.28,"developerDx":0.14,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Pair to `outbound-webhooks`. Confidence: high.","tags":["webhooks","inbound","events","agents"]},{"id":"deep-research-agents","slug":"deep-research-agents","name":"Deep research & multi-hop web agents","shortName":"Deep research","problemStatement":"Multi-step web research with citations, extractions sources, and agent-friendly APIs — beyond single-shot search.","whenNotTopPick":"Perplexity Sonar/Agent API already adopted, or Tavily/Exa fit simpler RAG search.","categoryIds":["cat-search-scrape"],"depotPickProviderId":"parallel","depotPickRationale":"Watch Perplexity Sonar deprecation timing. Confidence: medium-high on Parallel positioning.","runnersUpIds":["perplexity","exa","tavily","you-com"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Watch Perplexity Sonar deprecation timing. Confidence: medium-high on Parallel positioning.","tags":["research","search","citations","agents"]},{"id":"text-in-image-generation","slug":"text-in-image-generation","name":"Text-accurate image generation","shortName":"Text-in-image","problemStatement":"Generate marketing/UI images where spelled text on the image must be correct (posters, mockups, diagrams).","whenNotTopPick":"Vector/brand consistency (Recraft), Flux quality (BFL/Fal), or OpenAI already bundled.","categoryIds":["cat-media-gen"],"depotPickProviderId":"ideogram","depotPickRationale":"Media-gen specialty. Confidence: high on Ideogram typography reputation.","runnersUpIds":["recraft","black-forest-labs","fal","openai-media"],"weights":{"capabilityFit":0.32,"pricePerformance":0.12,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.12,"lockInRisk":0.08,"maturity":0.1},"methodologyNotes":"Media-gen specialty. Confidence: high on Ideogram typography reputation.","tags":["image","typography","design","media"]},{"id":"agent-push-notifications","slug":"agent-push-notifications","name":"Mobile push notifications for agents","shortName":"Agent push","problemStatement":"Notify humans on iOS/Android when agents need input or finish long jobs — without building APNs/FCM infrastructure.","whenNotTopPick":"Multi-channel orchestration (Knock/Courier/Novu) matters more than raw push.","categoryIds":["cat-comms"],"depotPickProviderId":"onesignal","depotPickRationale":"Comms. Confidence: high.","runnersUpIds":["knock","novu","courier"],"weights":{"capabilityFit":0.22,"pricePerformance":0.14,"latencyReliability":0.14,"developerDx":0.28,"privacyPosture":0.08,"lockInRisk":0.06,"maturity":0.08},"methodologyNotes":"Comms. Confidence: high.","tags":["push","mobile","notifications","hitl"]},{"id":"on-device-edge-batch","slug":"on-device-edge-batch","name":"Budget on-prem-ish batch open-model jobs","shortName":"Budget batch open models","problemStatement":"Run large offline batches on cheap distributed/consumer GPU networks when latency SLOs are hours not milliseconds.","whenNotTopPick":"Need predictable enterprise GPU (CoreWeave/Nebius) or TEE (Chutes).","categoryIds":["cat-gpu-cloud"],"depotPickProviderId":"salad","depotPickRationale":"Batch/cost focus. Confidence: medium — Salad price floor moves; pin /pricing.","runnersUpIds":["vast-ai","chutes","thunder-compute","verda"],"weights":{"capabilityFit":0.18,"pricePerformance":0.32,"latencyReliability":0.14,"developerDx":0.12,"privacyPosture":0.08,"lockInRisk":0.08,"maturity":0.08},"methodologyNotes":"Batch/cost focus. Confidence: medium — Salad price floor moves; pin /pricing.","tags":["batch","budget","gpu","open-models"]}]}