Google opens Jules coding agent to everyone

Most AI coding assistants act like a backseat driver, forcing you to watch every token generate in real time. Google wants to flip that script entirely.

On Wednesday, Google officially took Jules, its asynchronous AI coding agent, out of beta and launched it worldwide. Powered by Gemini 2.5 Pro, Jules takes a very different approach from the current crop of editor tools: you give it a prompt, and it works entirely in the background.

Code running while your laptop is closed

Instead of living inside your local IDE, Jules clones your GitHub repo into an isolated Google Cloud virtual machine. It plans the changes, edits the files, and runs the tasks asynchronously.

As Google Labs product director Kathy Korevec explained, Jules works like an extra set of hands. You can assign it an issue, shut down your computer, and walk away. Hours later, you can return to find the task completed with branches created and pull requests opened.

With the public launch, Google added several features requested during beta testing. Jules now supports Environment Snapshots to cache dependencies and install scripts, cutting down spin-up times. It also integrates directly with GitHub Issues, handles multimodal inputs, and works on completely empty repositories for greenfield builds.

Interestingly, data from SimilarWeb showed Jules pulled in 2.28 million visits during its preview, with 45% of traffic coming from mobile devices. Developers are literally using the web app on their phones to kick off automated fixes while on the move.

Paywalls and privacy fine print

As Jules exits preview, Google is putting up the expected guardrails and paywalls.

The free introductory plan now caps users at 15 tasks per day with three concurrent executions, down sharply from the 60 daily tasks allowed during beta. To do more, you have to subscribe to Google AI Pro ($19.99 a month) for 5x limits, or the top-tier Google AI Ultra ($124.99 a month) for 20x capacity. Eligible college students can snag a free year of the Pro tier.

Google also clarified its data privacy terms following developer questions. The company stated that data from public repositories may be used to train models, but code from private repositories will not be used for model training.

Why it matters

We think Jules marks a pivotal shift from synchronous autocomplete to true background delegation in software engineering.

Tools like Cursor and Windsurf are unbeatable when you want rapid inline pair-programming. But for mundane upkeep—like bumping library versions, writing boilerplate unit tests, or cleaning up messy prototype code—staring at an editor is a waste of time. Jules lets developers treat an AI agent like a junior colleague who works in the background while you focus on architecture.

Google's architectural choice here is clever. By tethering Jules to Google Cloud virtual machines, Google turns its massive infrastructure into a moat that smaller startups cannot easily replicate for free.

That said, asynchronous delegation requires immense trust. If a synchronous tool hallucinates, you catch it in two seconds. If an asynchronous agent goes down a rabbit hole in a VM, you might wait hours only to get a messy pull request. At 15 free tasks a day, it is definitely worth testing on repetitive maintenance chores, but moving your core workflow over will require paying up and keeping a very close eye on its PRs.