Google’s Jules Coding Agent Exits Beta, Joins the AI Coding Arms Race
After months in testing, Jules is now public—and it’s not just another Copilot clone. Google is betting on autonomous agents as the future of software development
Google’s rival to GitHub Copilot, OpenAI’s Codex, and Cursor is now publicly available. After three months in beta, the company’s AI coding assistant, Jules, has launched, now powered by Google’s Gemini 2.5 Pro model.
A product from Google Labs, Jules was initially an experimental AI agent first announced in December 2024. Six months later, it entered public beta. The company described it as “an early glimpse of what a true coding agent could become. Not a co-pilot, not a code-completion sidekick, but an autonomous agent that reads your code, understands your intent, and gets to work.”
Developers can connect Jules with their existing repositories, including GitHub. The agent will copy the codebase into a secure Google Cloud virtual machine, and can perform tasks including writing tests, building new features, generating audio changelogs, handling bug fixes, and bumping dependency versions.
Google’s coding agent was free to use, but that ends today. The company is introducing new pricing tiers, beginning with an entry-level plan for beginners and scaling up based on whether users subscribe to Google AI Pro or Ultra. Pro users will get 5x higher usage limits, which should cover most day-to-day coding needs. For more demanding, multi-agent workflows, the Ultra tier offers 20x the limits, optimized for power users and advanced development.
The company reveals that to date, Jules has helped “thousands” of developers complete “tens of thousands” of tasks, with over 140,000 code updates shared publicly. Based on user feedback, Google has given its agent a more user-friendly interface and has added multimodal support, meaning Jules can assess a web application and generate a visual representation of the results.
Jules’ launch follows closely on the heels of Google’s release of Opal, another coding agent, though with a different focus. While Jules is built to refine and enhance existing Python and JavaScript code, Opal generates it from scratch. Like Jules, Opal is starting as an experiment through Google Labs. In addition, Google also has a command-line agent called Gemini CLI that recently added support for GitHub. All of these bot innovations come at a time when interest in vibe coding has heated up, with some popular toolmakers, such as Cursor and Windsurf, being the subject of M&A conversations.