Walkthrough

This takes about half an hour, most of it waiting for the model. You need a GitHub repository connected as a project (see Getting started).

1. Turn on free models only. On Projects, switch on Free models only for your project. A lock appears next to its name.

2. Run an issue. Open an issue labelled ai-agent under Issues and click Run pipeline. Note the estimate under the model, tick Publish when done, and queue it.

3. Watch it. Open the job under Activity and follow the log, including the Readiness and AI review steps.

4. Check the run. When it finishes, open the run. Check the readiness score in the Task section and the Proof of work section.

5. Look at the pull request. On GitHub, the description has the same proof of work and Lyzentra's review of its change.

6. Ask for a change. Comment /agent fix with a small change. A Revise PR job runs, and a new commit and a reply appear on the pull request.

7. Teach it a rule. Comment something general, like /agent fix we always use Optional instead of null here. When it's done, Projects → Project rules shows a suggested rule. Approve it.

8. Review someone else's pull request. Open a pull request by hand, then comment /agent review (or use the AI review box on Projects).

9. Turn on automatic reviews. Switch on Review every new pull request and open another pull request. It gets reviewed without you asking.

10. Check a rule is followed. Add a rule you can see in the code, such as Add a short Javadoc comment to every new public class, then run another issue and look at the Changes tab.