Knowledge-graph / Graph-RAG memory
Build entity-relation memory so agents answer multi-hop questions better than flat vector RAG alone.
At a glance — Oct 2026
Updated 2026-10-27 · forward this page: /jtbds/knowledge-graph-rag#briefScores as of 2026-09-14 (43 days old; Depot marks scores stale after 30 days). The pick is still Depot's current editorial pick; it has not been re-reviewed since. Prices below carry their own dates.
- Depot Pick
- Cognee · 8.07/10 for this job (curated estimate, high confidence)
Confidence: medium — Cognee OSS/EU posture; confirm hosted SLA.
- Runner-up
- Weaviate · 8.26/10
Scores higher than the pick on: Capability fit, Latency / reliability, Maturity / track record.
- When not to use the pick
- Weaviate hybrid already enough, Zep temporal graphs fit sessions better, or Diffbot KG extract is the input stage only.
- Lock-in / migration
- The pick has lower stored lock-in risk than runner-up Weaviate (3 vs 3.5). Pick exit signals: open source (provider tag `oss`).
- Sourced price (no comparable unit)
Pick: no sourced price in this job's unit (no_sourced_price).
No cheapest SKU (no_comparable_unit_for_job). No price was invented.
- Free option
- No free resource lists this job. A cheap paid SKU is not a free option.
- Failure modes to watch
- Stale retrieval: Store the source URL and the time it was fetched. If a newer canonical page exists, or two stored sources disagree, do not answer from the older one.
- Context overflow: If the input exceeds the window, say so and retrieve a narrower slice. Do not answer from a truncated head as if it were the whole source.
- Policy conflict: If cited sources contradict each other, return the conflict and the URLs. Do not pick a winner silently.
- Ungrounded answer: Require a source URL or passage id on factual sentences. Drop or mark any sentence whose source does not contain the claim.
- Confirmation skip: Gate those tool names on an approval id recorded before the call. A missing approval is a failure even if the downstream tool returns success.
- Eval contamination: Keep gold labels out of the agent prompt. Store the harness version beside the score. Do not compare a number across a benchmark revision without saying the revision changed.
Cite: Agent Provider Depot, Depot Pick for knowledge-graph-rag: cognee, scores as of 2026-09-14. Agents: GET /api/decide?jtbd=knowledge-graph-rag
for support agents · for rag agents
Cognee
Confidence: medium — Cognee OSS/EU posture; confirm hosted SLA.
Pricing (directional): As of 2026-09-14 (cognee.ai/pricing): Free $0 (1 workspace, 1M tokens). Standard $1.00/1M tokens + $5/extra workspace/mo. Enterprise BYOC engagement (6–24 mo packages, startup discount). Apache-2.0 OSS with SQLite/LanceDB/Kuzu defaults. ECL…
Lock-in risk: higher = more lock-in (worse). Overall already inverts this dimension.
Graph memory when documents→ontology matter. Zep for temporal bipartite graphs; Mem0 for simplest hosted memory; Letta for stateful runtime (ex-MemGPT).
Quote for Knowledge-graph / Graph-RAG memory
Depot Pick stays Cognee. Prices below are sourced SKU rows only (as of 2026-10-27). Input and output are not blended. GET /api/quote?jtbd=knowledge-graph-rag
Nothing was compared. no comparable sourced unit for this job; chat input $/1M and other units were not borrowed pickIsPriceWinner is null.
Sourced price unavailable for the Depot Pick. No row in the pricing index.
4 of 4 candidates have no sourced price row. 4 are excluded from the Matching price comparison (none). no comparable sourced unit for this job; chat input $/1M and other units were not borrowed
Runners-up
Weaviate
Best when hybrid search and modularity matter. Cloud for convenience; self-host when lock-in anxiety spikes.
Zep
Pick when agents need temporal/entity graph memory, not just vector snippets. Mem0 for simpler recall.
Diffbot
Structured extract / KG peer. Firecrawl remains llm-ready-extract default; Diffbot when you need schema'd…
When NOT to use the top pick
Weaviate hybrid already enough, Zep temporal graphs fit sessions better, or Diffbot KG extract is the input stage only.
Full ranking (score table)
Ranked with this JTBD's weights. Scores are curated estimates unless marked otherwise. Agents: prefer /api/decide then this table for tie-breaks.
Score methodology for this job
Confidence: medium — Cognee OSS/EU posture; confirm hosted SLA.
capabilityFit: 32%pricePerformance: 12%latencyReliability: 14%developerDx: 12%privacyPosture: 12%lockInRisk: 8%maturity: 10%