docs
Documentation
Quickstart
Install the SDK, wrap your agent, verify your first signed trail. 10 minutes.
SDK & CLI reference
The whole surface on one page: Python, TypeScript, the verify CLI, env vars.
Evaluators
Score recorded trails against rubrics you write, with an LLM judge on models you choose.
The Improve loop
Cluster failures, propose patches, replay regressions, approve — every step recorded.
For AI agents
Written for agents: exact commands, zero signup, raw markdown at /docs/agents.md.
The loop
How observe, understand, and improve fit together — evaluators, replay, gates.
The Tail
What a loop record holds, how signing works, and how evidence export behaves.
The trail format spec (draft v0.1) ison GitHub. Watch the changelog.