Agent Runner

Agent Runner is a local workflow runner for coding agents.

It lets you define multi-step workflows - plan, implement, validate, fix, review, open a PR - and run those steps through the agent CLIs you already use: Claude Code, Codex, Copilot, Cursor, and OpenCode.

The key difference: the workflow lives outside the agent’s context window. The agent still reasons, writes code, and fixes problems. Agent Runner owns the sequencing, retries, state, validation loops, and resumption.

https://github.com/user-attachments/assets/e40c22f5-87dc-4158-9cc0-6bf9f6ede0ba

Quick start

macOS and Linux (Homebrew)

brew tap Codagent-AI/tap
brew install --cask agent-runner
agent-runner

First launch opens the guided setup and tutorial. The Homebrew install also installs Agent Runner’s helper CLIs, including agent-plugin and agent-validator.

Linux tarball fallback

If you are not using Homebrew, download the latest Linux tarball for your architecture from releases:

curl -LO https://github.com/Codagent-AI/agent-runner/releases/latest/download/agent-runner_linux_amd64.tar.gz
tar xzf…