what is repo-steward fr?

Repo Steward is an autonomous agent that takes over the day to day busywork of maintaining an open source project. Running on a schedule or triggered with a button press, it triages incoming issues, reviews pull requests over multiple rounds, joins repository discussions, writes bug fix pull requests, and watches the uptime of your project websites, all while keeping a live dashboard of what it has done and what it plans to do next. The project is built for maintainers whose time gets eaten up copying pull request diffs into a chat window to get an opinion on them. Instead of doing that by hand, the steward runs that same loop on its own across every repository you point it at, and shows its work openly rather than just reporting that something happened. A few design choices stand out. Everything the steward writes, like a review or a reply, sits in draft mode until you flip a live toggle, so nothing is posted publicly under your name until you trust it. The steward itself never merges or closes anything on its own judgment. Only a click or a typed decision from you executes an action, and that action runs under your own GitHub login because you were the one who decided. You can also type a plain sentence describing what you want done on any pending decision, and the tool interprets it and carries it out. Installing it requires a Linux machine with a systemd user session, Python 3, the jq command line tool, and the GitHub CLI logged in with push access to your repositories. You copy a configuration file, list your repositories and limits, and run an install script, after which a local dashboard is available in your browser showing decisions that need you, pull requests ready for a final look, and what the steward intends to do on its next scheduled run. By default it runs using Claude Code as its underlying AI engine, but it can be switched to other coding agent command line tools, including local models, at install time. The project is released under the MIT License.