Google launched its AI coding agent, Jules, out of the beta version on Wednesday, a little more than two months after its preview beginnings in May.
Propelled by Gemini 2.5 Pro, Jules is a coding tool based on asynchronous agents that integrate into Github, clones, code bases in Google Cloud Virtual Machines and use AI to correct or update code while developers focus on other tasks.
Google initially announced Jules as a Google Labs project in December and made it available to beta testers via a public overview during its conference on E / S developers.
Kathy Korevec, product director at Google Labs, told Techcrunch that the improved stability of the tool had led the decision to remove it from the beta version after receiving hundreds of updates from the user and quality interface during its beta phase.
“The trajectory of the place where we are going to give us a lot of confidence that Jules is there and will be long term,” she said.
With the wider deployment, Google introduced structured pricing levels for Jules, starting with a free “introductory access” plan capped at 15 individual and three simultaneous daily tasks, down compared to the limit of 60 tasks during the beta version. The paid levels of Jules are among the Google AI Pro and Ultra plans, at the price of $ 19.99 and $ 124.99 per month, and offer respectively to 5 × and 20 × higher subscribers.
Korevec has noted that the packaging and Jules’ prices are based on “real” information gathered in the past two months.
“The cap of 60 tasks helped us study how the developers use Jules and gave us the information we needed to design the new packaging,” she said. “The 15 / day is designed to give people an idea of knowing if Jules will work for them on real project tasks.”
Google has also updated Jules’ privacy policy to be more explicit about how it forms AI. If a repository is public, its data can be used for training, but if it is private, Korevec said that no data was sent.
“We have received some user comments that [the privacy policy] It was not as clear as we thought, and so we are most of that. We haven’t changed anything about what we do on the training side, but we have changed the language, “said Korevec.
During the beta version, Google said that thousands of developers had discussed tens of thousands of tasks, which has caused more publicly shared code improvements. The initial comments led the Google Labs team to add new capacities, in particular by reusing previous configurations for the execution of faster tasks, integrating the GitHub problems and taking charge of multimodal input.
So far, the two main users of Jules are AI fans and professional developers, said Korevec.
By performing asynchronously in a virtual machine, Jules is distinguished from the superior AI coding tools such as the cursor, the winsurf and the adorable, which all work synchronously and oblige users to look at the exit after each prompt.
“Jules works like an additional set of hands … You can mainly launch tasks, then you could close your computer and keep it away if you wish and then come back a few hours later. Jules would have these tasks carried out for you, compared to this session”, explained Korevec or use a synchronous agent, you would be linked to this session, “said Korevec.
This week, Jules received a deeper integration with Github to automatically open traction requests – just as it could open the branches – and a functionality called momentshots Environment to record the outbuildings and install the scripts as a snapshot for a faster and more coherent task execution.
From room coding to mobile use, beta trials have informed the development of Jules
Since entering public beta, Jules has recorded 2.28 million visits worldwide, including 45% from mobile devices, according to data from the similar market provider, examined by Techcrunch. India was the best traffic market, followed by the United States and Vietnam.
Google did not share any details based on users of Jules and its large geographies.
Korevec told Techcrunch that during the beta version, the team observed that many people used jules of traditional mood coding tools to correct the bugs that could have been implemented or extend the project coded by the atmosphere to make it more ready for production.
Originally, Jules forced users to have an existing code base. But Google quickly realized that many potential users – such as those who try other AI tools – might want to explore it without one. Korevec said that the company quickly allowed Jules to work even with an empty repository. This has helped increase its scope and use.
The Google Labs team has also noticed an increasing number of users accessing Jules via their mobile devices. Although the tool does not have a dedicated mobile application, Korevec said users access it via its web application.
“As it is a great case of use that we see emerging, we absolutely explore the features that people need a lot on mobile,” she noted.
In addition to beta testers, Korevec said Google already uses Jules to help develop certain projects internally, and there is now a “big push” to use the tool on “many more projects” in the company.
