Auto-Patch & Agent Autonomous
Manual AI fixes, GitHub commits, and optional autonomous patching.
Two ways to fix issues
Manual (Security tab)
Pick an issue in Issue List → View Fix → review AI suggestion → Commit Correction on GitHub. Best for one-off fixes you want to review first.
Agent Autonomous (settings)
Background runner on a schedule (default every 5 minutes): locate → AI → apply via GitHub. Configure at Settings → Integrations → [your integration] → Agent Autonomous.
Manual workflow
- Connect GitHub (Settings → Integrations → GitHub)
- Open the integration → Security → select an issue
- View Fix — AI uses stack trace + repo locate when possible
- Review the patch → Commit Correction
- Verify in your repo and deploy
Agent Autonomous settings
- Enabled — turn the runner on/off
- Mode — Suggest only (prepare fixes, no auto-apply) or Autonomous (pipeline runs locate → AI → apply)
- Severity threshold — only issues at or above this level
- Max patches per day — daily cap
- Schedule — realtime, hourly or daily batch (cron still checks every 5 min when enabled)
- Run now — trigger one cycle immediately from the settings page
Requires a linked GitHub repo and a paid plan for AI apply.
Where issues come from
The autonomous runner and manual fix flow use the same issue sources as the Security tab:
- SDK errors and vulnerabilities (noise-filtered)
- Agent-ranked findings (performance, instability, spikes)
- Not TestingAI headless noise (those stay on the TestingAI tab)
Prerequisites
- GitHub repository connected with write access
- PRO plan for AI suggestions and apply
- Issues visible in Issue List (SDK events) or detected by the agent
Need help? Join our Discord.