Google's AI Coding Agent Jules Exits Beta with New Pricing
Google has officially moved its AI coding agent, Jules, out of beta, just over two months after its public preview debut in May. Powered by the advanced Gemini 2.5 Pro, Jules is designed as an asynchronous, agent-based coding tool that integrates seamlessly with GitHub. Its core function involves cloning codebases into Google Cloud virtual machines, where its AI then autonomously fixes or updates code, freeing developers to concentrate on other tasks.
First unveiled as a Google Labs project last December, Jules was made accessible to beta testers during the I/O developer conference. Kathy Korevec, director of product at Google Labs, indicated that the decision to transition Jules out of beta was driven by significant improvements in its stability, a result of hundreds of UI and quality updates implemented throughout its testing phase. Korevec expressed strong confidence in the tool’s future, stating, “The trajectory of where we’re going gives us a lot of confidence that Jules is around and going to be around for the long haul.”
With this wider rollout, Google has introduced a structured pricing model for Jules. An “introductory access” free plan allows for up to 15 individual tasks per day and three concurrent tasks, a reduction from the 60-task limit during the beta period. For more intensive use, Jules is available through Google AI Pro and Ultra plans, priced at $19.99 and $124.99 per month, respectively, offering subscribers five times and twenty times higher task limits. Korevec explained that these packaging and pricing decisions were informed by “real usage” insights gathered over the past few months, with the beta’s 60-task cap serving as a crucial data collection point for understanding developer behavior. The 15-task daily limit for the free tier is intended to help users evaluate Jules’s efficacy on real-world projects.
Addressing user feedback, Google also updated Jules’s privacy policy to provide greater clarity on how it handles data for AI training. While data from public repositories may be used for training, Korevec assured that no data is transmitted from private repositories. She clarified that the policy language was refined to be more explicit, rather than altering the underlying data handling practices.
During its beta phase, Jules saw widespread adoption, with thousands of developers tackling tens of thousands of tasks, leading to over 140,000 publicly shared code improvements. Initial feedback proved invaluable, prompting the Google Labs team to introduce new capabilities such as reusing previous setups for faster task execution, deeper integration with GitHub issues, and support for multimodal input. Korevec noted that Jules’s primary users thus far include both AI enthusiasts and professional developers.
Jules distinguishes itself from other popular AI coding tools like Cursor, Windsurf, and Lovable through its asynchronous operation. Unlike these synchronous tools, which demand real-time user attention to output after each prompt, Jules runs in the background within a virtual machine. This allows developers to initiate tasks and then disengage, returning later to find the work completed. As Korevec put it, “Jules operates like an extra set of hands… you can basically kick off tasks to it, and then you could close your computer and walk away from it if you want and then come back hours later.”
Recent enhancements to Jules include a deeper integration with GitHub, enabling the tool to automatically open pull requests, similar to how it could previously open branches. A new feature called “Environment Snapshots” allows users to save dependencies and install scripts, facilitating faster and more consistent task execution.
Beta trials also offered unexpected insights into user behavior. Data from market intelligence provider SimilarWeb, reviewed by TechCrunch, indicates that Jules logged 2.28 million visits worldwide during its public beta, with a notable 45% of access occurring via mobile devices. India led in traffic, followed by the U.S. and Vietnam. While Google did not share specific user base numbers or top geographies, Korevec highlighted observations that many users leveraged Jules to refine “vibe-coded” projects—less formal, experimental coding—either to fix bugs or make them production-ready. Initially, Jules required an existing codebase, but Google quickly adapted it to work with empty repositories, significantly broadening its appeal and usage. The surprising prevalence of mobile access via the web app, despite no dedicated mobile application, has prompted Google to actively explore features tailored for this emerging use case. Beyond external users, Google itself is increasingly deploying Jules internally, with a “big push” to integrate the tool into a growing number of its own development projects.