Google’s AI coding agent Jules is now out of beta

On Wednesday, Google officially unveiled its AI coding assistant, Jules, moving it out of beta just over two months after its initial public preview in May. Built on the advanced Gemini 2.5 Pro architecture, Jules is designed as an asynchronous coding agent that seamlessly integrates with GitHub, allowing developers to clone codebases into Google Cloud virtual machines. The tool harnesses AI capabilities to fix or update code while developers focus on their primary tasks. Kathy Korevec, director of product at Google Labs, shared that the decision to launch Jules officially was influenced by enhanced stability and numerous user interface improvements made during the beta phase. "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," she stated. The rollout comes with new structured pricing tiers for Jules. An introductory free plan allows up to 15 individual daily tasks and three concurrent tasks, a reduction from the 60-task limit offered during the beta phase. For users seeking more extensive capabilities, paid tiers are available as part of the Google AI Pro and Ultra plans, priced at $19.99 and $124.99 per month, respectively, offering significantly increased task limits. Korevec explained that these pricing structures are informed by actual usage data from beta testers. "The 15 tasks per day is intended to help users assess whether Jules can effectively assist them with real project tasks," she noted. Additionally, Google has updated Jules’ privacy policy to clarify its data handling practices. While data from public repositories may be used for AI training, no data is sent from private ones. Korevec emphasized that the changes were made in response to user feedback regarding clarity. During its beta testing, Jules supported thousands of developers in completing tens of thousands of tasks, resulting in over 140,000 publicly shared code improvements. Feedback from early users led to enhancements such as the ability to reuse previous setups for quicker task execution and support for multimodal input. Jules distinguishes itself from other AI coding tools by operating asynchronously. This means users can initiate tasks and then step away, returning later to find their requests completed, unlike traditional synchronous agents that require constant monitoring. Recent updates also include deeper integration with GitHub for automatic pull requests and a new feature called Environment Snapshots, which saves dependencies and installation scripts for quicker task processing. Data from SimilarWeb indicates that since its beta launch, Jules has attracted 2.28 million visits globally, with a notable 45% of traffic coming from mobile devices. India leads in user engagement, followed by the U.S. and Vietnam. Although initially requiring users to have an existing codebase, Google quickly adapted Jules to work with empty repositories, broadening its appeal. As more users access Jules through mobile web apps—despite the absence of a dedicated mobile application—Korevec affirmed that the team is exploring necessary features for mobile users. In addition to external testers, Google has begun using Jules to enhance its internal projects, with ambitions to deploy the tool across more initiatives moving forward.