AgentRail Docs

AgentRail is the control plane that runs your coding agents — Claude Code, Codex, or Cursor. It is not a coding agent itself and never writes code. It gives the agent you already use durable repo context, bounded execution, independent review, and verification gates, so its work is cheaper, inspectable, and safe to leave unattended.

The primary flow is the hosted dashboard + self-hosted runner: sign in to the dashboard, connect your GitHub repos, then run agentrail login and agentrail runner on your machine. The runner claims queued issues over HTTP, executes them locally with your own LLM keys, opens PRs, and reports back. Login is required to use the CLI (only login, logout, whoami, and --help are exempt).

Start here

• Getting Started — what AgentRail is, how to install it, and a quickstart from sign-up to a reviewed PR.
• Core Concepts — how the context engine, bounded execution, review gates, AFK mode, memory, and skills actually work.
• CLI Reference — every agentrail command, grouped by what you're doing.

Going further

• Configuration — the .agentrail config, choosing an agent, and bringing your own LLM keys.
• Dashboard & hosted runner — sign in, connect GitHub, queue issues, and run the local worker.