Runs
Every plan, reply and implementation Rookery ran on your repos, newest first. Select a run to see its log, step by step.
To start a run, add the label rookery:trigger to an issue on GitHub. Rookery posts a plan there. Comment approve on the issue, and it writes the code and opens a draft pull request. Getting started
What Rookery does on each issue in this repo, stage by stage. Each stage has an agent, which does the work, and can have skills, the how-to guides that the agent follows. The defaults work as they are: change them only if you want to. How the workflow works
- Drag an agent or a skill from the shelf onto a stage.
- Check the bar that shows at the bottom of the page.
- Open a pull request. Nothing changes until someone merges it.
Options for this repo, and the notes that Rookery keeps about it. Both are on Rookery's server, not in your repo: a change here applies to the next run, with no pull request.
A skill is a how-to guide that an agent follows, for example how to write the tests first. These skills come with Rookery. None is on by default: you add one to a stage of a repo's workflow.
- To use a skill, select a repo on the left, come back here, and select Add to … on the skill. Or drag it onto a stage in the repo's Workflow tab.
- A skill with the same name in your repo (.rookery/skills/ or .claude/skills/) or in <owner>/.rookery replaces the one here.
Add your own API key, and Rookery uses it for every run on the repos of your own GitHub account. How keys work
- The selected key pays for the runs, and its provider receives your code.
- Rookery keeps each key encrypted, uses it only for your runs, and never shows it again.
- The daily caps still apply.
