On Wednesday, Google launched its beta AI coding agent, Jules, just two months after its public preview debut in May.
Powered by Gemini 2.5 Pro, Jules is an asynchronous, agent-based coding tool that integrates with GitHub and integrates with Crones CodeBases and Google Cloud Virtual Machines, which uses AI to modify or update code while focusing on other tasks.
Google originally announced Jules as a Google Labs project in December, making it available to beta testers through a public preview of the I/O Developer Conference.
Kathy Korevec, product director at Google Labs, told TechCrunch that the improved stability of the tool has encouraged the decision to remove it from beta after receiving hundreds of UI and quality updates in the beta stage.
“The trajectory we go gives us confidence that Joule has been around for a long time,” she said.
Google has introduced a structured price tier for Joules with a wider rollout. This started with a free “Introductory Access” plan that caps 15 individual daily tasks and three simultaneous tasks, starting with the beta version of the 60 task limit. Jules’ paid tiers are part of Google AI Pro and Ultra Plans, with prices of $19.99 and $124.99 per month, with 5x and 20x limits, respectively.
Korevec noted that Joule’s packaging and pricing is based on “actual use” insights gathered over the past few months.
“60 Task Cap provided us with the information we needed to study how developers use Joules and design new packages,” she said. “15/day is designed to make people feel like Jules will work for them on actual project tasks.”
Google has also updated its Jules privacy policy to make it even clearer how to train AI. If the repository is public, that data may be used for training, but in private, Korevec said that no data will be sent.
“We’ve had a bit of feedback from users who say (the privacy policy) wasn’t as clear as I thought. So, most of the time, we didn’t change anything about what we’re doing on the training side, but we changed the language,” Korevec said.
During the beta, Google said thousands of developers were working on tens of thousands of tasks, resulting in over 140,000 code improvements being made public. Early feedback has led the Google Labs team to add new features. This includes reusing previous setups to make task execution faster, integrate with GitHub issues, and support multimodal input.
The two main users of Jules so far are AI enthusiasts and professional developers, Korevec said.
By running asynchronously on a virtual machine, Joules are separated from cursors, windsurfs and beloved Top AI coding tools. All of these should work synchronously and watch the output to the user after each prompt.
“Jule behaves like a set of extra hands… You can basically kick off tasks, and close the computer if you want to and come back in a few hours. Jule is tied to that session if you’re doing it with a local agent, or if you’re using a sync agent,” explains Korevece.
This week, Jules will receive a deeper integration with GitHub, automatically opening pull requests, receiving a feature called environment snapshots, and install the script as a snapshot for faster, more consistent task execution.
From atmospheric coding to mobile use, Beta Trial informed the development of Joules
Since entering the public beta, Joule has recorded 2.28 million visits worldwide. That 45% is from mobile devices for each Market Intelligence Provider SigrayWeb data reviewed by TechCrunch. India was the top market for transportation, followed by the US and Vietnam.
Google did not share details about Jules’ user base and its top geography.
Korevec told TechCrunch that during the beta, many people observed that during the past atmosphere coding tool Joules would use traditional atmospheric coding tools to fix bugs that could have been implemented or to extend vibe-coded projects to make them production-ready.
Originally, Joule requested that users have an existing codebase. But Google quickly realized that many potential users might want to explore one by one, like those trying out other AI tools. Korevec said the company is now able to work with Joules in empty repositories. It helped to increase its range and usage.
The Google Labs team also noticed an increasing number of users accessing Joules via mobile devices. The tool does not have a dedicated mobile app, but Korevec said users are accessing it via a web app.
“We’re seeing a massive use case, so we’re exploring more of the features people need on mobile,” she pointed out.
Korevec says, alongside the beta tester, Google is already using Joules to develop several projects internally, and there is now a “big push” using the tool in the company’s “more projects”.
