Google Jules
Also known as: Jules, Jules by Google, Google Jules agent
Google Jules is an asynchronous AI coding agent by Google, powered by Gemini 2.0. Unlike real-time pair programmers, Jules works autonomously in the background on GitHub issues and pull requests. You assign it a task (bug fix, feature, refactor), Jules clones your repo into a sandboxed cloud environment, implements the changes, runs tests, and opens a pull request for your review. Jules is in public beta as of early 2026.
Agent Information
- Provider
- Supported IDEs
- GitHub (web), Browser
Feature Support Matrix
| Feature | Support Level | Details |
|---|---|---|
| Execution | ||
| Automatic Context Awareness | ✅ Full Support | Jules clones the entire repository into a sandboxed cloud environment for each task, giving it full awareness of the cod... |
| Filesystem Access | ✅ Full Support | Jules has full filesystem access to the cloned repository in its sandboxed cloud environment. It can create, edit, and d... |
| MCP Server Support | ❌ No Support | Jules does not support MCP (Model Context Protocol) servers. It operates in an isolated cloud sandbox and does not conne... |
| Editor Integration | ||
| Broad IDE Integration | ❌ No Support | Jules operates entirely through a web interface at jules.google and through GitHub. It has no IDE plugin or extension fo... |
| Model Support | ||
| Claude 4 Support | ❌ No Support | Jules is powered exclusively by Google's Gemini 2.0 model family. It does not support Anthropic's Claude models. |
| Claude 3 Support | ❌ No Support | Jules uses Google's Gemini 2.0 models only. Anthropic Claude models are not supported. |
| Context Window Management | ✅ Full Support | Jules ingests the full repository into its context using Gemini 2.0's large context window (up to 1M tokens). This enabl... |
| Debugging | ||
| Console Error Integration | ❌ No Support | Jules works asynchronously in a cloud sandbox and has no browser console integration. It can read test output and termin... |
| Interactive Element Selection | ❌ No Support | Jules is an asynchronous agent working in a cloud sandbox. It has no visual UI element selection capability. |
| Live Web Preview | ❌ No Support | Jules operates in a cloud code execution environment. It can run tests and build steps, but does not provide a live web ... |
| Configuration | ||
| Dedicated Instruction File | ⚠️ Partial | Jules reads repository files including README and documentation as part of its context. There is no official named instr... |
| Fine-Grained Instruction Control | ❌ No Support | Jules does not have a file-pattern or directory-level instruction system. All task instructions are provided in the task... |
| Supports Scoped Instructions | ❌ No Support | Jules does not support scoped or hierarchical instruction systems. Instructions are provided per-task in the task descri... |
| Planning | ||
| Planner Strategy | ✅ Full Support | Jules generates a step-by-step implementation plan before writing code, which is shown to the user for approval. Only af... |
Frequently Asked Questions
Jules is an asynchronous AI coding agent that works on GitHub issues and tasks in the background. You assign Jules a task — such as fixing a bug or implementing a feature — and it independently clones your repository into an isolated cloud environment, writes the code, runs tests, and opens a pull request for your review. You review and merge the PR when satisfied.
Jules is powered by Google's Gemini 2.0 model family. It uses Gemini 2.0 Flash for efficient task execution and can leverage Gemini 2.0 Pro for complex reasoning tasks.
Jules is currently in public beta (as of early 2026) and is free to use during the beta period. Pricing for general availability has not been announced. Access is available via jules.google with a Google account.
Jules integrates with GitHub and can work with any public or private GitHub repository you grant it access to. It clones your repo into a sandboxed cloud environment for each task, so it does not run on your local machine.
Jules is fundamentally asynchronous — you do not interact with it in real time. You assign a task and come back later to review a pull request. Tools like Copilot and Cursor work synchronously in your IDE as you type. Jules is better suited for delegating entire self-contained tasks rather than pair programming.
Jules excels at self-contained, well-defined tasks: fixing bugs from issue descriptions, implementing small features, writing or updating unit tests, updating dependencies, and performing code migrations. Tasks requiring interactive decisions or ambiguous requirements are better handled by synchronous agents.
Yes. Jules runs the repository's existing test suite as part of its task execution in the sandboxed environment. It uses test results to validate its implementation and may iterate on the code if tests fail.
Jules operates in a sandboxed cloud environment and does not have access to external services, live databases, or production systems. It works only with the code and tests in your repository.
As of early 2026, Jules is in public beta with access open globally, though it initially launched in the United States. Availability may vary. Check jules.google for the latest access information.
Jules understands task descriptions written in multiple languages, as Gemini 2.0 is a multilingual model. However, the Jules interface and documentation are primarily in English.