Timeline

Google makes its Jules coding agent generally available

Google's answer to OpenAI's Codex and Cognition's Devin: an agent that works on a copy of your repository in the cloud and returns finished changes, rather than autocompleting in the editor.

  • Models & capabilities
  • Minor

Google moved its Jules coding agent out of public beta and made it available to anyone, with a free tier and paid limits tied to its Google AI Pro and Ultra subscriptions. Jules is “asynchronous”: rather than suggesting completions as a developer types, it clones a copy of the user’s code repository into a cloud virtual machine, works through a task on its own — reading the code, forming a plan, making and testing changes — and comes back with a finished set of edits to review. First shown in December 2024 and opened to a public beta at Google I/O in May 2025, it is powered by the company’s Gemini 2.5 Pro model.

The design placed Jules in a specific and increasingly crowded category. Its closest comparisons were not editor assistants like Cursor or GitHub Copilot but the cloud agents that had appeared through 2025: OpenAI’s Codex, launched three months earlier, and, further back, Cognition’s Devin. All shared the same premise — that a coding tool could be handed a whole task and left to run, with the human stepping in to check the result rather than to steer every keystroke.

Google reported that during the beta thousands of developers had run tens of thousands of tasks, producing more than 140,000 publicly shared code changes. The general-availability release added an improved interface, reuse of previous environment setups, integration with GitHub issues, and, Google said, hundreds of bug fixes. It was a second front in the company’s coding push, arriving weeks after it had shipped the open-source Gemini CLI for the terminal — evidence that, like its rivals, Google now treated coding as a category worth attacking from several directions at once.