Know what your agents did —
and prove it later
One .wrap() on your chain or agent — signed events, linked outcomes, replay in the Explorer.
Cryptographically prove what your agents knew and did — at decision time.
No credit card · Demo mode works without an API key · Free tier to start
Drop-in tracers for the stacks you already run. Same decision → outcome model, same Explorer — whether you're on LangChain, LlamaIndex, or multi-agent frameworks.
AudifactLangChainTracer · AudifactLlamaIndexTracer · AudifactTracer — all via .wrap()
Prove what happened —
at the moment of the decision
Hash-linked events capture inputs, tool steps, and outputs when they matter — not reconstructed from logs days later.
One line to trace
tracer.wrap(your_chain) — no logging pipeline, no custom middleware. Ship today, audit tomorrow.
Privacy you control
Hashes and metadata by default. Turn on include_context=True only when you need full inputs, docs, or tool outputs.
Decisions with receipts
Every run becomes a linked decision → outcome pair. Readable public_decision labels in the Explorer — not a wall of JSON.
Pick your starting point
Runnable demos, copy-paste snippets, and use-case guides — from weekend projects to enterprise RAG.
Notes bot & weekend research
Simplest LangChain wrap. Demo mode, no API key. Perfect for learning and personal RAG.
See the basic guide →LCEL chains + tool agents
Move fast with confidence. Debug agent runs, show investors clean event trails.
See the startup guide →LlamaIndex retrieval + agents
QueryEngine, structured Pydantic outputs, and tool-using agents with full provenance.
See LlamaIndex guide →LangGraph decision workflows
Multi-step graphs with node-level audit trails for high-stakes agentic workflows.
See LangGraph guide →Start small. Scale when it matters.
Run examples locally in demo mode tonight. Add an API key when you're ready to persist signed events to the cloud.