Kit · 2026-10-03
Agent kits
One page per agent class. Each kit reuses the current Depot Pick, the job quote, the free index, failure modes, and fetched case studies. It does not rescore and it does not invent a price. Call GET /api/kit for the index, then GET /api/kit?class={id}. A blank class is 400. An unknown class is 404. Agents that cannot spend call /api/kit?class={id}&cost=free and must not treat the paid pick as free. Any other cost value is 400.
Coding agent
Edits a repository, runs commands in a workspace, and checks the result against tests or a diff.
Research agent
Gathers sources for an open-ended question and returns an answer that names those sources.
Browser agent
Completes a task by reading and operating a website in a browser, then stops before a consequential submit unless confirmation exists.
Computer-use agent
Operates a desktop or mobile GUI, not only a single site API, and checks that the control it used is the one it intended.
Voice agent
Hears or speaks on a call or a realtime channel, and confirms account-changing actions before they run.
Support agent
Answers a customer from the operator's policy and escalates when sources disagree or the action is consequential.
Data and ETL agent
Moves or reshapes records between files, tables, and documents, and reports rows that did not match the schema.
RAG agent
Answers from a retrieved corpus and says when the corpus does not support the claim.
Multi-agent orchestrator
A lead assigns subtasks to workers and accepts a result only when it matches the assignment.
Eval and judge agent
Scores a trace or a patch against a rubric, without the gold label in its prompt, and records the harness version.
Scheduled and ops agent
Runs a job on a queue or a clock, is safe to retry, and alerts when the job or its traces go dark.
Personal assistant
Acts for one person across sessions, keeps that person's memory separate, and waits before sending or booking.
Sales and outreach agent
Drafts outreach from sourced fields about a person or account, and does not send until a person approves.
Security review agent
Reads a change or a trace for secrets, overly broad permissions, and untrusted text copied into a tool prompt. It does not produce an attack.
Local and offline agent
Runs without calling a hosted model API, using local weights, a local index, or an air-gapped serve path.