Overview

The Coord Runner is a desktop app that executes agentic jobs on your machine. It connects to your Coord workspace, picks up jobs that are ready, and runs AI agents locally to complete the work.

Why you need a runner

Agentic jobs require an AI agent to execute them. The Coord Runner provides the execution environment — it clones repositories, starts the agent, manages the work, and reports results back to Coord.

Without a runner, agentic jobs stay in Ready and are never picked up.

How it works

1. The runner connects to your Coord workspace and watches for jobs that are ready.
2. When a job becomes available, the runner picks it up.
3. The runner prepares the environment — for repository jobs, it clones the code and creates a branch.
4. The runner starts an AI agent (Claude Code, Codex, or Gemini CLI) with the job instructions.
5. The agent works on the task. You can watch progress in real time, answer questions, and approve tool calls from the runner or the Coord web app.
6. When the agent finishes, the runner packages the results and updates the job in Coord.

For repository jobs, the runner automatically commits, pushes to GitHub, and opens a pull request. You do not need to handle any of this manually.

The runner interface

The Coord Runner has three main tabs:

• Status — shows active running jobs and controls for starting or stopping the runner.
• Logs — streams real-time output from the agent as it works, so you can follow along.
• History — lists completed jobs with their outcomes, so you can review past work.

Supported platforms

Automatic retries

If the agent encounters a transient error — a crash, timeout, or network hiccup — the runner automatically retries the work. It picks up where it left off using the previous session context. You only see a job as Failed if all automatic retries are exhausted.

Graceful stop

You can stop a running job gracefully from the runner. When you request a stop, the agent finishes its current operation and exits cleanly rather than being terminated immediately. This gives the agent a chance to save its progress.

Multiple runners

A workspace can have multiple runners, and a single runner can serve multiple workspaces. This makes it easy to scale execution across a team.

Next steps

• Installation — download and install the runner
• Setup and sign-in — connect the runner to your workspace
• Configuration — fine-tune runner settings