GatwAI

Google Jules

Google Jules for asynchronous GitHub tasks: cloud VMs, daily limits, private-repository data rules and proposed maintenance workflows.

Provider sources checked

What does Google Jules do?

Jules takes engineering tasks from a GitHub repository and works on them in a cloud VM. The workflow covers a task, plan, changes and a reviewable diff or pull request. This suits work where every intermediate step does not need to be followed together inside an editor. Its distinction from an IDE assistant is asynchronous execution: a task can continue while the developer does something else. Tests, documentation and bounded dependency upgrades are especially easy to describe. Working setup instructions and realistic acceptance criteria matter. Evaluate value through accepted work, including review and necessary corrections.
Editorial research · 2026-11-14How we research →

Who is this tool for?

  • Developers with GitHub projects and clear maintenance tasks.
  • Teams that first review hosted execution and account terms.

Strengths

  • Cloud execution runs tasks independently of an open local editor.
  • Plans and diffs support concrete acceptance review.
  • GitHub integration fits existing issue and PR workflows.

Limitations and requirements

  • Setup scripts need terminating commands; persistent development servers are unsupported there.
  • Daily allowances and concurrency limit task throughput.
  • Paid access currently targets individual Google accounts according to plan documentation.

Who is behind the product?

Company
Google Labs / Google
Headquarters
Not separately established by the reviewed sources.
Company founded
Not separately established by the reviewed sources.
Product launch
Introduced in December 2024; public beta on 20 May 2025.

Google dates the broader public release out of beta to 6 August 2025. Some FAQ wording still says beta, so the dated announcement is treated separately. Company founding and product history are different facts.

Suggested ways to use it

Suggested workflows or verification steps. These are not hands-on tests; for unverified products they are not a recommendation to use the service.

Suggested: tests for a known defect

  1. Record a reproducible defect and expected result in an issue.
  2. Select repository and setup, then inspect the plan.
  3. Review the test and patch in the PR, including unexpected changes.

Goal: A traceable maintenance proposal.

Suggested: bounded version upgrade

  1. Name the library, target version and incompatible changes.
  2. Delegate installation, edits and appropriate tests to Jules.
  3. Check dependencies and runtime behavior before merging.

Goal: A reviewed update candidate rather than an indiscriminate upgrade.

Pricing, allowances and data

What should you check before choosing a plan?

The current product page lists 15 daily tasks on the free tier, 100 with Google AI Pro and 300 with Ultra. Daily limits use a rolling 24-hour window; allowances and model access can change. Regional subscription prices appear in Google's AI offering. A plan does not guarantee successful completion of every task.

What should you check about data and permissions?

The FAQ explicitly states that private repository content is not used for model training. Code is nevertheless copied to and processed in an internet-connected cloud VM. Restrict repository permissions and keep secrets out of project files. The private-repository statement is not a blanket assurance about all usage or feedback data.

External reviews and available evidence

No GatwAI hands-on test. This profile relies on current primary sources rather than an independent original test of today's version.

Reported positives

  • The documented plan-and-PR workflow provides inspectable intermediate and final outputs.

Reported cautions

  • The suggested workflows are editorial application ideas. Provider examples do not establish general time savings.

Product development and notable changes

  • Public release out of beta

    Google announced public release and larger Pro and Ultra allowances; today's limits are listed on the current product page.

    Source for this change ↗

Which alternative fits which task?

OpenAI Codex →

For local and cloud tasks with a different provider and model offering.

Frequently asked questions

Must my computer stay on during a task?

Jules works in a cloud VM and can continue after you leave the app.

Can every Google Workspace account simply upgrade?

Do not assume so. Current plan documentation identifies individual Google accounts and still-limited organizational upgrade paths.