Jules is Google's autonomous coding agent, powered by Gemini, that connects to a GitHub repository, plans an approach to a task you describe, writes a code diff, and opens a pull request for you to review, rather than working inside your editor turn by turn.
Who builds it
Jules is built by Google, using Gemini models to plan and generate code. It's designed for asynchronous work: you hand off a task like a bug fix, a dependency upgrade, or a framework migration, and Jules works on it in the background while you do something else.
Core features
- GitHub-native workflow: you select a repository, describe the task, and Jules produces a plan, then a diff, then a pull request.
- Autonomous planning: Jules works out an approach to the task itself using Gemini before writing any code, rather than requiring you to specify every step.
- Background/async execution: tasks run without blocking your own work, and you review the resulting diff and PR when it's ready.
- Typical use cases: bug fixes, dependency and framework version updates, and general feature or test work delegated as a self-contained task.
Pricing
Jules is offered in three tiers, distinguished by daily task volume, concurrency, and model access rather than separate feature sets:
- Jules (Free): 15 tasks per rolling 24-hour window, 3 concurrent tasks, running on Gemini 2.5 Pro — positioned by Google for evaluating Jules on real work.
- Jules Pro: 100 tasks per rolling 24-hour window, 15 concurrent tasks, with higher access to the latest model starting with Gemini 3 Pro; unlocked through a Google AI Pro subscription rather than sold as a separate Jules-only plan.
- Jules Ultra: 300 tasks per rolling 24-hour window, 60 concurrent tasks, with priority access to the latest model starting with Gemini 3 Pro; unlocked through a Google AI Ultra subscription.
- Paid tiers currently require an individual Google Account (for example, a @gmail.com address); Google says enterprise and Workspace account support is in active development.
Data privacy
Google has said Jules is private by default for private repositories: prompts, diffs, and commits are used only to execute that session and are not used to train Google's models, and no data from a private repo is sent for training. Public repositories are treated differently — their data may be used for training — so the practical privacy guarantee depends on whether the connected repository is public or private.
Who it's for
Developers and teams on GitHub who want to hand off well-defined, self-contained coding tasks — a dependency bump, a bug fix, a small migration — and get back a reviewable pull request without staying in the loop step by step. The free tier suits solo evaluation, while Pro and Ultra fit developers who already pay for Google AI Pro/Ultra and want Jules' higher daily task volume and concurrency as part of that subscription.