Docs

Reelier records what your AI agent does once, then replays it deterministically — 0 tokens, milliseconds, a receipt every run. It’s built for agents that do the same job over and over: the same booking lookup, the same status check, the same weekly report.

The engine is open source (AGPL-3.0) and runs on your machine. Reelier Cloud is an optional, opt-in place to push your run receipts so they accrue in one dashboard instead of living only in local .reelier/ files.

Start here

• Quickstart

  Install, init, run, push — the 60-second path.

• How it works

  Record, compile, replay, self-heal. The five atoms. The escalation ladder.

• Skills (SKILL.md)

  The file format a recorded workflow compiles into.

• Recording

  How to capture a real agent session with the MCP recording proxy.

• Cloud

  Sign in with GitHub, push receipts, schedules, self-host instead.

• Models (BYOK)

  Bring your own key. Any OpenAI-compatible endpoint, or native Anthropic.

• CLI reference

  Every command, one line each.

• FAQ

  AGPL, privacy, what can and can't be replayed, is my key safe.