Google releases AI agent Jules for programming

Google has released its AI programming agent, Jules, after a beta testing period that began in May. The tool uses the Gemini 2.5 Pro model and allows you to run tasks asynchronously, integrates with GitHub, creates environments on Google Cloud virtual machines, and updates code without developer participation in real time, TechCrunch reports.

Jules was announced as a Google Labs project in December and has undergone hundreds of UI and quality updates during testing. According to Google Labs product director Katie Korevek, stability improvements were the main reason for leaving beta.

With the release, the company also introduced a new pricing structure. The free plan allows for up to 15 tasks per day, 3 of which can be done simultaneously. The paid plans — AI Pro and Ultra — cost $19.99 and $124.99, respectively, and offer 5x and 20x higher limits.

Google also updated its privacy policy: public repositories can be used to train models, but private ones cannot. Jules also gained deeper integration with GitHub after leaving beta: it can now open pull requests, just like branches, and a new Environment Snapshots feature saves dependencies and scripts to speed up tasks.

During the beta, users launched tens of thousands of tasks and shared over 140,000 public code improvements. The tool learned to work even with empty repositories, making it more accessible. According to SimilarWeb, since May, 2.28 million users have visited the Jules website, 45% of which are from mobile devices. The largest traffic was recorded from India, the United States, and Vietnam.

Despite the lack of a dedicated mobile app, Jules is widely used via the web interface, and Google is considering additional smartphone support. There is also a growing number of projects within the company that are already using the tool.