Ready for Agent
A local harness that turns issues labeled ready-for-agent into reviewed, merged pull requests using coding agents like OpenCode, Codex, Grok Build, and Claude Code.
At a Glance
Self-hosted harness under the MIT license, run locally on your own machine. Requires your own coding agent (external cost, not sold by this tool).
Engagement
Available On
Listed Oct 2026
About Ready for Agent
Ready for Agent is an open-source harness by Berend de Boer that takes issues marked ready-for-agent in GitHub, GitLab, Azure DevOps, or Linear and drives a coding agent to implement them, review the code, open a pull request, and merge when allowed. It runs locally against an existing git clone and is started through npm. The latest listed release is v1.0.0.
What It Is
Ready for Agent is an agent harness: a deterministic loop around a coding agent CLI. You add a local repository, label issues ready-for-agent (or tag an Azure Boards work item), and the issues appear in a web UI at a local address. The author describes it as automating the implementation and review stages of the workflow Matt Pocock describes, while the human handles the grilling session, specification, and ticket creation.
How the Workflow Runs
For each issue the harness creates a fresh worktree, installs packages, and asks a headless agent to implement the issue. A separate review phase follows, ideally using a different model, then a pull request is created and GitHub status checks are awaited. Depending on the per-repo Merge Policy (Off, Classify, or Always), the PR is merged automatically or left for a human in an attention lane. You can pick a parent issue to implement all its child issues, or choose "Implement locally" to inspect the work before any commit or PR exists.
Agents, Forges, and Trackers
Four interchangeable backends are supported: OpenCode 2, Codex, Grok Build, and Claude Code. Build and review models and effort can be set globally or overridden per repository. Supported forges are GitHub (via gh), GitLab (via glab), and Azure DevOps (via a personal access token); Linear can serve as the issue tracker for GitHub repositories. It optionally integrates with keymaxxer for secrets handling.
Setup Path
Prerequisites are git, the forge CLI for your repositories, and at least one authenticated coding agent. Local agent execution requires Linux (x64 or arm64) with cgroup v2 and a delegated systemd user manager (systemd 254+); macOS and Windows packages can start the app but refuse local execution. The tool does not create issues itself and does not pick issues automatically; you start each implementation.
Community Discussions
Be the first to start a conversation about Ready for Agent
Share your experience with Ready for Agent, ask questions, or help others learn from your insights.
Pricing
Open Source (MIT)
Self-hosted harness under the MIT license, run locally on your own machine. Requires your own coding agent (external cost, not sold by this tool).
- MIT License
- Runs on your local laptop, avoiding cloud costs
- Supports GitHub, GitLab, Azure DevOps, and Linear issues
- Works with OpenCode, Codex, Grok Build, or Claude Code (authenticated separately)
Capabilities
Key Features
- Turns ready-for-agent labeled issues into merged pull requests
- Interchangeable backends: OpenCode 2, Codex, Grok Build, Claude Code
- Per-repo agent and model overrides
- Separate build and review phases with configurable models and effort
- Per-repo Merge Policy: Off, Classify, Always
- Implement locally option to inspect work before a PR
- Implement a parent issue with all its child issues
- Kanban-style UI with attention lane for human review
- Supports GitHub, GitLab, Azure DevOps, and Linear issues
- Runs locally against an existing clone
- Jump command to continue an agent session in terminal or tmux
- Optional keymaxxer secrets integration
Integrations
Demo Video

