Jules

Jules is Google's asynchronous coding agent: instead of watching an assistant type in your editor, you assign it a task and walk away. Jules clones your GitHub repository into an isolated Google Cloud virtual machine, drafts a plan you can review, makes the changes, and opens a pull request for you to review, revise or merge.

Key facts: Jules began as a Google Labs experiment in December 2024, entered public beta at Google I/O in May 2025 and became generally available on 6 August 2025, when Google reported beta users had already shared more than 140,000 code improvements publicly. You can assign work from the Jules web app or CLI, or simply add a “jules” label to a GitHub issue, and it can also scan a codebase for TODOs, fix failing GitHub Actions checks on its own pull requests and run scheduled tasks.

Plans: Free allows 15 tasks per rolling 24 hours with 3 running at once. Jules in Google AI Pro ($19.99/month) raises that to 100 tasks a day and 15 at once, and Jules in Google AI Ultra raises it further to 300 tasks a day and 60 at once, with higher or priority access to the latest Gemini model. Google says Jules does not train on private repository content.

Best for: developers with a backlog of well-scoped GitHub issues, bug fixes and chores that can be reviewed as pull requests. Watch out: it is GitHub-only with no editor integration, and paid upgrades currently require a personal Google account, so Workspace and enterprise buyers should check eligibility before planning around it.

Pros

• Fully asynchronous: assign a task and walk away instead of watching an assistant type in your editor.
• Drafts a plan you can review, then makes the changes and opens a pull request for review, revision or merge.
• Assign work from the web app, the CLI, or simply by adding a jules label to a GitHub issue.
• Also scans codebases for TODOs, fixes failing GitHub Actions checks on its own pull requests, and runs scheduled tasks.

Watch out for

• It is GitHub-only with no editor integration, so non-GitHub workflows are out of scope.
• Paid upgrades currently require a personal Google account, so Workspace and enterprise buyers should check eligibility first.
• Pull requests still need human review; async does not mean merge-blind.

Best for

Developers with a backlog of well-scoped GitHub issues, bug fixes and chores that can be reviewed as pull requests. Who should skip it: developers who want real-time, in-editor pair programming or who do not use GitHub.

Pricing

Free allows 15 tasks per rolling 24 hours with 3 running at once. Jules in Google AI Pro ($19.99/month) raises that to 100 tasks a day and 15 at once, and Jules in Google AI Ultra raises it further to 300 tasks a day and 60 at once, with higher or priority access to the latest Gemini model. Check the official pricing page for current pricing.

FAQs

How is Jules different from an in-editor assistant? It is asynchronous: you assign a task, it works in an isolated Google Cloud virtual machine, and returns a pull request.

How do I assign work? From the Jules web app or CLI, or by adding a jules label to a GitHub issue.

What is the free allowance? 15 tasks per rolling 24 hours with 3 running at once, per the page.

Does Jules train on my private repositories? Google says Jules does not train on private repository content.