Jules coding agent from Google now available

Google has officially launched Jules, an autonomous asynchronous coding agent, for general availability. Previously in beta, Jules is now accessible globally anywhere the Gemini model is available, with no waitlist. Jules is powered by Gemini 2.5 Pro and was first introduced in December 2024 via Google Labs as a preview of what a fully autonomous coding agent could become, not just a co-pilot or autocomplete tool, but a system capable of reading code, understanding developer intent, and executing tasks independently.

Jules coding agent from Google now available: An asynchronous AI coding agent, powered by Gemini 2.5

During the beta period, thousands of developers completed tens of thousands of tasks using Jules, resulting in over 140,000 publicly shared code improvements. Based on feedback from these developers, Google refined the user interface, resolved hundreds of bugs, and introduced several new features, including:

• The ability to reuse previous setups for faster task execution
• GitHub Issues integration
• Multimodal input support
   

Jules functions asynchronously and integrates directly with a user’s existing repositories. It securely clones the codebase into a dedicated Google Cloud virtual machine (VM), then analyzes the project context to perform a variety of coding tasks, including:

• Writing and updating tests
• Building new features
• Fixing bugs
• Bumping dependency versions
• Generating audio changelogs of commit history
   

Upon completion of a task, Jules presents a summary of its plan, its reasoning, and a diff of the changes it made. Jules is designed to be private by default, it does not train on private code, and user data remains isolated within its execution environment. With Gemini 2.5 Pro at its core, Jules is equipped to reason across large codebases and handle concurrent tasks within the cloud VM system. This combination enables Jules to manage complex, multi-file changes at scale. Key capabilities include:

• Real-world codebase integration - No sandbox required; it operates directly within live projects
• Parallel execution - Multiple tasks can run simultaneously in the cloud
• Transparent workflow - Users can see the agent’s reasoning and proposed plan before any code changes are applied
• GitHub integration - Jules functions directly within GitHub workflows
• User steerability - Users can modify Jules’ plans before, during, or after execution
• Audio summaries - A narrated changelog is generated to help developers understand recent updates
   

With its public launch, Google has introduced structured access tiers for Jules:

• Introductory access - Basic functionality suitable for exploring Jules across different projects
• Google AI Pro - Includes 5x higher usage limits, designed for daily developer workflows
• Google AI Ultra - Offers 20x higher limits, tailored for large-scale, multi-agent coding operations
   

These access levels are now being rolled out to Google AI Pro and Ultra subscribers. College students who qualify can also access AI Pro for free for one year. Full usage limits are detailed at jules.google.