Shepherd PRs to merge automatically
Bots check PRs every 30 minutes, send failures back to a cloud agent, review clean PRs and auto-merge high-confidence, low-risk ones.
Done withGrok Bot
- What they did
- Each engineer bot keeps a shared Notion database of PRs. Every 30 minutes it checks each PR for Bugbot or security comments, failing CI and merge conflicts. Problems go back to the cloud agent, and the row moves to Working. Clean PRs are marked Ready for Review and get a review run. Highly confident, low-blast-radius PRs are merged automatically. The author reviews the rest.
- How it went
- The author finds tasks ready to merge almost every morning. The fleet manages over 200 cloud agents at once, up from about 15 by hand. The source gives no merge-rate figures.
- Worth knowing
- The author says rapid transcript-checking routines, like the 5-minute P0 mode, burn tokens faster than expected, so keep them for true urgency.
Try it yourself with Grok Bot
Every 30 minutes, check open pull requests in [repo]. Send failing PRs back to a coding agent to fix, and review the clean ones. Only auto-merge a PR if you are highly confident and it is low-risk, such as [examples]; otherwise ping me with a summary. Pause and ask me before merging anything that touches [sensitive areas].
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: x.ai · Undated
Related use cases
Plan and book a short-notice work trip
Grok Bot researched flights and booked the one Vince picked, though he had to take over the computer to fill in data and found it harder than booking himself.
Analyze YouTube thumbnails from channel analytics
Grok Bot connected to YouTube analytics compared top-video thumbnails' CTR against the channel average and offered new thumbnail ideas.
Build four games in parallel with agents
Four Grok 4.20 agents built a tower defense, mini Factorio, CHIP-8 emulator and roguelike shooter in parallel over a day.
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Hand narrow code changes to Devin for pull requests
A team used Devin for a month on small fixes like chart formatting and API edge cases, with tests and PRs; a complex full-stack task went poorly.
Build a booking and payment app for tattoo artists
A former programmer used Replit Agent over a weekend to build a booking app with calendar, Google OAuth, Stripe deposits and notifications, though time zones and UI needed back-and-forth.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.