Jules
An autonomous coding agent from Google Labs that handles bug fixes, tests, version bumps, and feature building by cloning your repo in a Cloud VM and creating GitHub PRs.
At a Glance
About Jules
Jules is an experimental autonomous coding agent built by Google Labs, powered by Gemini 3 Pro. It integrates directly with GitHub, clones your repository into a Cloud VM, and executes coding tasks — from bug fixes to feature builds — without requiring you to stay in the loop. The tool is currently available via the web at jules.google.com and also accessible through a CLI and API.
What It Is
Jules is an async coding agent designed to handle the coding tasks developers don't want to do: bumping dependency versions, writing or running tests, fixing bugs, and building new features. Rather than acting as an inline autocomplete tool, Jules operates autonomously in a sandboxed Cloud VM, generates a plan for your review, makes code changes, and then opens a pull request on GitHub. The workflow is designed to let developers hand off a task and move on to other work while Jules completes it in the background.
How the Workflow Works
Jules follows a four-step process:
- Task assignment: Select a GitHub repo and branch, write a detailed prompt, or use the
juleslabel on a GitHub issue to assign tasks directly from GitHub. - Planning: Jules fetches the repo, clones it to a Cloud VM, and uses Gemini 3 Pro to generate a step-by-step plan. You review and approve the plan before any code is changed.
- Execution: Jules modifies files, runs existing tests or creates new ones, and verifies changes work inside the VM.
- PR creation: Jules produces a diff for review and opens a pull request on GitHub for you to approve and merge.
Jules also reads an AGENTS.md file from the root of your repository, which can describe agents, tools, input/output conventions, and other project-specific context to help Jules generate more relevant plans.
Platform and Access
Jules is accessible as a web application, a CLI, and a developer API, making it usable both interactively and as part of automated workflows. The CLI reference and API documentation are published at jules.google/docs. The Google Labs team also maintains a public "Jules Awesome Prompts" repository on GitHub (google-labs-code/jules-awesome-list) with curated prompt examples covering everyday dev tasks, debugging, documentation, testing, package management, and more — the repo had over 3,000 stars as of mid-2025.
Underlying Technology
Jules is built on Gemini 3 Pro, Google DeepMind's model, and developed by the AIDA team at Google Labs. The Cloud VM environment means Jules can install dependencies, run build scripts, and verify that changes actually work before presenting them to the developer — not just generate code in isolation.
Update: Gemini 3 Pro and Current Availability
Jules is currently live and available at jules.google.com. The homepage notes that Jules is "brought to life with Gemini 3 Pro," indicating a recent model upgrade from earlier Gemini 2.5 Pro access (which remains available on the base plan). The tool is described as experimental, consistent with its Google Labs origin. The awesome-list repository was created in May 2025 and last updated in May 2026, signaling active community and product development.
Community Discussions
Be the first to start a conversation about Jules
Share your experience with Jules, ask questions, or help others learn from your insights.
Pricing
Jules
Get started with real coding tasks.
- 15 tasks per day
- 3 concurrent tasks
- Powered by Gemini 2.5 Pro
Google AI Pro
Professional plan with Everything in For Individuals and More generous rate limits for power users.
- Everything in For Individuals
- More generous rate limits
- Flexible AI credit pool
Google AI Ultra
Google AI Ultra plan with Everything in Google AI Pro and More generous rate limits.
- Everything in Google AI Pro
- More generous rate limits
- Flexible AI credit pool
Capabilities
Key Features
- Autonomous coding agent
- GitHub integration with PR creation
- Cloud VM code execution and verification
- Plan review before code changes
- Bug fixing
- Test generation and execution
- Version bumping
- Feature building
- AGENTS.md support for repo context
- Browser notifications for task completion
- CLI access
- API access
- GitHub issue label assignment (jules label)
- Concurrent task execution
- Diff review interface
Integrations
Demo Video

