Google Launches "Jules": Its AI Coding Agent

Google officially announced the launch of "Jules," its advanced AI-powered coding agent, shortly after its initial unveiling in May. Powered by the Gemini 1.5 Pro model, Jules operates as an asynchronous, agent-based coding tool that seamlessly integrates with the GitHub platform. It clones code repositories to virtual machines on the Google Cloud, using AI to fix or update code while developers remain focused on other tasks.

Google had initially announced Jules as a project within its Google Labs in December, making it available to testers in a public beta during the I/O developer conference. Kathy Corvik, Product Director at Google Labs, explained that the tool's remarkable stability was the primary reason for its official launch, after receiving hundreds of updates related to the user interface and performance quality during the beta phase.

With its expanded launch, Google revealed various pricing plans for Jules, starting with a free "introductory access" tier that allows for 15 individual tasks per day and a maximum of three concurrent tasks, a reduction from the 60-task limit available in the beta. Paid plans are part of the Google AI Pro and Ultra tiers, costing $19.99 and $124.99 per month respectively, granting subscribers twenty-five times higher usage limits.

“Pricing Image” — Source: Pixabay. License: CC0.

Corvik noted that Jules's packages and prices were determined based on "actual usage" insights gathered over the past two months. She added: "The 60-task limit helped us study how developers used the tool and provided us with the necessary data to design the new packages." She continued: "The 15-task per day limit is designed to give users a clear idea of whether Jules will meet their needs in their real-world projects."

Google also updated Jules's privacy policy to clarify how AI models are trained more transparently. If a repository is public, its data may be used for training purposes, but if it is private, Corvik confirmed that no data from it is sent.

“Privacy Policy” — Source: Pixabay. License: CC0.

During the beta period, Google stated that thousands of developers used Jules to process tens of thousands of tasks, resulting in over 140,000 publicly shared code improvements. Initial user feedback led the Google Labs team to add new features, such as reusing previous settings to speed up task execution, integration with "GitHub Issues," and support for multimodal inputs.

“Guitar Performance” — Source: Pixabay. License: CC0.

Jules Features and Integrations

According to Corvik, Jules's primary users so far are AI enthusiasts and professional developers. By operating asynchronously in a virtual machine, Jules distinguishes itself from leading competing coding tools like Cursor, Windswept, and Lovable, which all operate synchronously and require the user to monitor outputs after each request.

Corvik said: "Jules acts as an extra hand... You can simply assign it tasks, then close your computer and leave if you want, and come back hours later. You will find that Jules has completed those tasks for you, unlike what happens when using a local or synchronous agent, where you are tied to that session."

This week, Jules gained deeper integration with GitHub, allowing it to automatically open pull requests – in addition to its ability to open branches – and a feature known as "environment snapshots" was added, which saves dependencies and installs scripts as a single snapshot for faster and more consistent future task execution.

Jules's Spread and Evolution

Since Jules entered the public beta phase, it recorded 2.28 million global visits, 45% of which came via mobile devices, according to data from market intelligence firm SimilarWeb, which was reviewed by TechCrunch. India topped the list of visits, followed by the United States and Vietnam.

Google did not share specific details about Jules's user base or geographical distribution. However, Corvik told TechCrunch that during the beta, the team noticed many users transitioned to Jules from traditional coding tools to fix bugs or expand their projects and make them more production-ready.

Initially, Jules required an existing codebase. However, Google quickly realized that many potential users – such as those exploring other AI tools – might want to try it without this requirement. Corvik explained that the company quickly enabled Jules to work even with an empty repository, which contributed to expanding its scope of use.

The Google Labs team also noticed an increasing number of users accessing Jules via their mobile devices. Although the tool does not have a dedicated mobile app, Corvik confirmed that users were easily accessing it through its web application.

In addition to external testers, Corvik stated that Google is already using Jules to assist in the development of some of its internal projects, and that there is now a "strong trend" to adopt the tool in "many projects" within the company.