Skip to main content
AgentRail is a local-first control plane for coding agents. It sits beside your agent (Claude Code, Codex, Cursor, or a custom harness) and provides a single source of truth for every stage of the development lifecycle: assigned work, PR state, CI results, review feedback, real-time events, and ship decisions. You install the CLI, run agentrail init, and your agents stop polling four different APIs.

Quick Start

Install the CLI and get AgentRail running in under five minutes.

Task Lifecycle

Understand how issues move from intake to shipped code.

SDK Reference

Use the TypeScript or Python SDK to integrate AgentRail into your harness.

API Reference

Explore every public endpoint with request and response examples.

​
How it works

AgentRail normalizes provider issues (GitHub, Linear) into compact tasks, routes them to the right agent, and keeps track of every lifecycle transition so your agent never has to guess what to do next.
1

Install and initialize

Install the CLI with npm install -g @agentrail-core/cli and run agentrail init to create local config and your first agent profile.
2

Start the server

Run agentrail server start to bring up the local API and wake configured agents automatically.
3

Connect providers

Run agentrail provider connect github (or circleci / linear) to link your issue tracker and CI provider.
4

Verify with doctor

Run agentrail doctor to confirm the API is reachable, your agent credentials work, and assigned work is visible.

​
Choose your path

Claude Code / Codex / Cursor

Copy-paste agent instructions for managed local runners.

Custom harness (TypeScript)

Build an external harness or MCP server using the TypeScript SDK.

Custom harness (Python)

Build a workflow runner or internal service using the Python SDK.

Self-hosting

Run AgentRail on your own hardware or a VPS with Docker.