Asynchronous coding meets structured pricing in Google’s long-term vision for AI-assisted development
From preview to public release
Just over two months after its public preview debut at Google I/O in May, Jules — Google’s AI coding agent powered by Gemini 2.5 Pro — has officially launched out of beta. The tool, originally announced as a Google Labs project in December, is designed to work asynchronously, running in Google Cloud virtual machines while developers focus on other tasks.
During beta, Jules received hundreds of UI and quality updates, boosting stability enough for Google to greenlight a full release. “The trajectory of where we’re going gives us a lot of confidence that Jules is around for the long haul,” said Kathy Korevec, director of product at Google Labs.
How Jules works differently
Unlike synchronous coding assistants such as Cursor, Windsurf, and Lovable, Jules operates like an extra set of hands. Developers can assign tasks, close their laptops, and return later to find them completed — freeing them from session-bound interactions.
Key capabilities now include:
- GitHub integration for automatic pull requests and branch creation.
- Environment Snapshots to save dependencies and install scripts for faster, consistent execution.
- Support for empty repositories, expanding use cases beyond existing codebases.
- Multimodal input and GitHub Issues integration for more flexible workflows.
Pricing based on real-world usage
With the public launch, Google introduced structured pricing tiers informed by beta data:
- Introductory Access (Free) – 15 daily tasks, 3 concurrent tasks (down from 60 during beta).
- AI Pro ($19.99/month) – 5× higher limits.
- AI Ultra ($124.99/month) – 20× higher limits.
“The 60-task cap helped us study how developers use Jules… The 15/day is designed to give people a sense of whether Jules will work for them on real project tasks,” Korevec explained.
Privacy clarity, not policy change
Responding to beta feedback, Google updated Jules’ privacy policy for transparency:
- Public repositories – Data may be used for AI training.
- Private repositories – No data is sent for training.
No underlying data practices changed — only the language for clarity.
Beta learnings: mobile growth and vibe coding
Beta trials revealed several surprising trends:
- Mobile use surge – 45% of visits came from mobile devices, despite no dedicated app.
- Top traffic markets – India, U.S., and Vietnam led adoption.
- Vibe coding adoption – Many used Jules to clean up or productionize experimental projects.
These insights have prompted Google to explore mobile-optimized features for future updates.
Internal adoption at Google
Beyond public users, Google is already employing Jules for internal projects, with a “big push” to expand its use across more engineering teams.
The bigger picture
Jules’ asynchronous model stands out in a crowded AI coding space, appealing to developers who want hands-off automation instead of step-by-step prompting. The move out of beta signals Google’s confidence not only in the technology but in a long-term market role — especially as mobile adoption and empty-repo onboarding open the tool to a wider audience.
