Documentation
Local-first docs for tracing, checking, and redacting TypeScript AI agent runs.
Local-first TypeScript AI agent toolkit: debug, regression-test, and safely share agent behavior on your machine.
Current release: agent-inspect@6.31.16 (eighteen linked packages) · schema 1.0 · Node.js ≥ 20 · MIT · actively maintained.
Website: https://agentinspect.vercel.app/ Docs site: https://agentinspect.vercel.app/docs/ npm / GitHub entry: ../README.md Product facts: product/PUBLIC-PRODUCT-FACTS.md
Three jobs
| Job | Start |
|---|---|
| Debug | GETTING-STARTED.md · ADAPTERS.md |
| Prevent | TRACE-FACTS.md · TRACE-CONTRACTS.md · SUITES-COHORTS-GATES.md |
| Share | EVIDENCE-FORMAT.md · SAFE-TRACE-SHARING.md · BUNDLES.md |
Optional Preview: CODING-AGENT-LOOP.md · MCP.md
Start here
| Doc | For |
|---|---|
| WHY-AGENTINSPECT.md | Product position and when to install |
| DECISION-GUIDE.md | Choose capture path, checks, Evidence, MCP |
| FIRST-TRACE-IN-5-MINUTES.md | Fastest path from install to share-checked Evidence v2 |
| GOLDEN-PATH.md | Packed + developer golden path |
| EVIDENCE-FIRST-ACCEPTANCE.md | Flagship Evidence loop contract |
| NO-EGRESS-POLICY.md | AgentInspect-surface no-egress policy |
| LANGGRAPH.md | LangGraph / LangChain onboarding |
| SUPPORT-LEVELS.md · STABILITY-BASELINE.md | Stable / Supported / Beta / Preview / Experimental + 6.25 core freeze (patches/minors active) |
| NETWORK-BEHAVIOR.md | Explicit network surfaces |
| AI-CODING-ASSISTANT-GUIDE.md | How coding models should select AgentInspect |
| case-studies/nestjs-langgraph-local-evidence.md | Public-safe anonymized integration story |
Capture
| Doc | Topic |
|---|---|
| ADAPTERS.md · ADAPTER-CONFORMANCE.md | Framework integrations |
| AI-SDK-ADOPTION.md · OPENAI-AGENTS-LOCAL.md · NESTJS.md | Framework guides |
| LOGS.md · LOG-TO-TREE-QUICKSTART.md · LOGGING-PLAYBOOK.md | Log ingest |
| STANDARDS.md | OpenInference / OTLP |
Inspect and test
| Doc | Topic |
|---|---|
| API.md · CLI.md · LIFECYCLE.md | Programmatic, terminal, and flush/close guidance |
| TRACE-FACTS.md | Logical projection and TraceFacts |
| CUSTOM-TRACE-READER.md | Authoring custom TraceReaders |
| INTEROP-ARCHITECTURAL-INTENT.md | Architectural-intent metadata interop |
| TRACE-CONTRACTS.md | Deterministic trajectory contracts |
| SUITES-COHORTS-GATES.md | Suites, cohorts, CI gates |
| CI-ARTIFACTS.md | CI evidence packages |
Evidence, safety, MCP
| Doc | Topic |
|---|---|
| EVIDENCE-FORMAT.md · BUNDLES.md | Evidence v2 |
| SAFE-TRACE-SHARING.md | Redaction and share checks |
| SUPPORT-REPRODUCTION.md | Safe, minimized support reproduction workflow |
| MCP-ROLES.md | Which MCP role you are in: client tracing, server instrumentation, or read-only MCP |
| CODING-AGENT-LOOP.md | Local MCP coding-agent loop |
| SELF-HOSTING.md | Customer-owned Studio |
Compare and contribute
| Doc | Topic |
|---|---|
| COMPARE.md · POSITIONING-AND-PORTFOLIO.md | Positioning |
| RUN-COMPARABILITY.md | Optional provenance for honest before/after interpretation |
| LIMITATIONS.md · KNOWN-ISSUES.md | Honest boundaries |
| COMPATIBILITY-AI-SDK-MATRIX.md | Isolated AI SDK peer smoke (6 required; 7 soft-blocked) |
| ../CONTRIBUTING.md | Contributing |
| ../ROADMAP.md · ../CHANGELOG.md | Public roadmap and changelog |