Google’s AI Coding Agent Jules Exits Beta: Now Available
▼ Summary
– Google launched its AI coding agent, Jules, out of beta, powered by Gemini 2.5 Pro, integrating with GitHub and using AI to automate code fixes.
– Jules transitioned from beta after hundreds of UI and quality updates, with pricing tiers now introduced, starting with a free plan capped at 15 daily tasks.
– Google updated Jules’ privacy policy to clarify that public repository data may be used for training, while private repository data remains unused.
– Jules operates asynchronously, allowing developers to offload tasks and return later, distinguishing it from synchronous AI coding tools like Cursor and Windsurf.
– During beta, Jules saw 2.28 million visits globally, with mobile usage at 45%, prompting Google to explore mobile-friendly features and broader internal adoption.
Google has officially launched its AI-powered coding assistant Jules out of beta, marking a significant milestone for the tool just two months after its public preview debut. Built on Gemini 2.5 Pro, Jules operates as an asynchronous coding agent, allowing developers to offload tasks like bug fixes and updates while they focus on other work. The tool integrates seamlessly with GitHub, cloning repositories into Google Cloud virtual machines for efficient AI-driven code improvements.
Originally introduced as a Google Labs experiment last December, Jules entered public beta testing during the company’s I/O conference in May. According to Kathy Korevec, Google Labs’ product director, the decision to exit beta was driven by hundreds of refinements that enhanced stability and usability. “The feedback and improvements gave us confidence that Jules is here to stay,” she noted.
With the general release, Google introduced tiered pricing for the service. A free plan now allows 15 daily tasks, down from the beta’s 60-task limit, while paid tiers under Google AI Pro ($19.99/month) and Ultra ($124.99/month) offer significantly higher task capacities. Korevec explained that the revised limits reflect real-world usage patterns observed during testing, ensuring developers can evaluate the tool’s effectiveness before committing.
Privacy concerns were also addressed with clearer policy language, though Google confirmed no changes to how data is handled. Public repositories may contribute to AI training, but private code remains untouched. “We refined the wording based on user feedback, our practices haven’t changed,” Korevec clarified.
During beta testing, Jules processed tens of thousands of tasks, generating over 140,000 public code improvements. User input led to new features like GitHub issue integration, multimodal input support, and Environment Snapshots, which save dependency configurations for faster task execution. Unlike competitors such as Cursor or Windsurf, Jules operates independently, users can assign tasks and return later to completed work without monitoring progress in real time.
Interestingly, 45% of Jules’ 2.28 million beta visits came from mobile devices, despite lacking a dedicated app. India led global traffic, followed by the U.S. and Vietnam. Korevec highlighted unexpected use cases, like developers refining “vibe-coded” projects or experimenting with empty repositories, prompting Google to expand accessibility.
Internally, Google has already adopted Jules for select projects, with plans to scale its use company-wide. As Korevec put it, “Jules acts like an extra pair of hands, you set it in motion and come back to results.” With deeper GitHub integration and mobile optimization on the horizon, the tool aims to redefine how developers collaborate with AI.
Image Credits: Jules / X
(Source: TechCrunch)
