Jules is Google's async coding agent for bounded GitHub work: bug fixes, dependency bumps, test additions, migrations, and follow-up maintenance tasks that can be described, planned, and reviewed as a pull request. The default workflow is user-started — choose a repo and branch, write a prompt, review the plan, then let Jules run in a Google Cloud VM. Jules also supports GitHub issue label triggers, Scheduled Tasks, and opt-in Suggested Tasks for repositories where proactivity is explicitly enabled.
Built on Google's Gemini model family, Jules uses repository context to draft changes across multiple files, follow project conventions, and present a reviewable diff. The free Jules plan lists 15 tasks per day and 3 concurrent tasks on Gemini 2.5 Pro; Google AI Pro raises that to 100 tasks/day and 15 concurrent tasks with latest-model access starting with Gemini 3 Pro; Ultra lists 300 tasks/day and 60 concurrent tasks. The async model lets developers start work, move on, and return to a plan or PR-shaped result.
Jules integrates with standard Git workflows by creating branches and pull requests for proposed changes after the plan/review loop. It is best suited to GitHub-centric teams that want async help on routine maintenance, small bug fixes, test coverage, and version migrations without keeping an IDE session open. Teams with self-hosting requirements, non-GitHub version control, or strict data-residency constraints should account for the Google Cloud VM architecture before adopting it.
