August 7, 2025
5 min read
Mels Dees
Google launches Jules, an AI coding assistant running on Gemini 2.5 Pro, enabling asynchronous code tasks and integration with GitHub.
Google launches AI coding agent Jules
Google has officially launched its AI coding assistant Jules and taken it out of beta. The tool, introduced as a public preview at the end of May, is now available as a regular product. Jules runs on Gemini 2.5 Pro and is designed as an asynchronous, agent-based programming assistant that supports developers in various tasks. It integrates with GitHub, clones repositories to a virtual cloud environment within Google Cloud, and independently performs operations on the code. This allows developers to outsource multiple processes without needing to be actively involved in every step. Unlike traditional AI tools that often require synchronous, direct interaction, Jules performs tasks completely asynchronously. Developers can initiate tasks, shut down their laptops, and return later to retrieve the results. This makes Jules particularly suitable for long-running or routine tasks that do not require continuous monitoring. During its beta phase, Jules supported thousands of developers. It was used for detecting and resolving bugs, expanding existing projects, and performing maintenance work based on previously generated code. Jules proved especially effective for repetitive development tasks such as creating test scripts, adjusting dependencies, and performing structural refactorings. Additionally, the agent offers support in generating changelogs and can read these logs aloud, enhancing usability in scenarios where visual attention is limited. With the official rollout, Google introduced a modified pricing structure. The free entry-level version allows users to perform fifteen tasks per day, with a maximum of three simultaneous processes—a significant reduction from the sixty tasks per day available during beta. For more intensive use, Pro and Ultra subscriptions are offered as part of Google’s broader AI suite. In Europe, these subscriptions are priced at €21.99 and €274.99 per month, respectively, providing five and twenty times more processing capacity, both in task volume and parallel sessions. This pricing is based on data collected during the test period. The original limits helped Google understand actual usage behavior to make targeted product positioning decisions. The current free version limits aim to give developers a realistic view of how Jules can be used in practical projects. Privacy terms have been tightened and clarified. Public repositories may be used to train the AI, while private projects are strictly excluded from this process. Private data is processed via Google Cloud but is neither analyzed nor shared. However, since every command runs through Google’s infrastructure, some developers may remain cautious despite explicit data minimization guarantees. During beta, Jules generated thousands of code improvements from tens of thousands of tasks. Valuable insights from this phase included the need to reuse previous setups for faster execution, support multimodal input, and deeper GitHub integration, including automatic pull request creation. The platform was also enhanced to work better without an existing codebase, allowing users to start even with an empty repository. This significantly improves accessibility, especially for users experimenting with AI without a current project. Usage statistics show Jules has registered millions of visits worldwide, with a large share coming from mobile devices. Despite lacking a native mobile app, many developers access the platform via its web interface on phones. Internally, Google now uses Jules for its own development projects and views it as a strategic building block within its development stack.Source: Google launches AI coding agent Jules - Techzine