Cursor's New Cloud Agents Write Code, Run Tests, and Send You Video Demos

Robot autonomously writing code at a desk while a human developer watches from a distance

Cursor, the AI-powered code editor, just launched its most ambitious feature yet: autonomous cloud agents that can clone your repository, write code, run tests, and deliver pull requests complete with screenshots, logs, and video demos of the finished work. The announcement is trending on X with over 5,800 posts and millions of views.

What Cursor's Cloud Agents Do

The new agents operate entirely in the cloud on secure virtual machines. You can delegate tasks from the desktop app, the web, Slack, GitHub, or Linear. The agent then clones your repo, edits code, runs tests, and iterates until everything works. When it's done, you get a pull request with full documentation.

The headline feature? Video demos. The agents don't just send you code diffs — they actually use the software they build and record video demonstrations of their work. As Cursor put it: "Demos, not diffs."

Already Powering 30% of Internal Merges

This isn't just a prototype. Cursor says these agents are already responsible for over 30% of their own internal code merges. That's a significant portion of a software company's engineering output being handled by AI agents rather than human developers.

Available on Pro plans, the agents are designed for tasks that can be delegated and reviewed asynchronously — bug fixes, feature additions, test writing, and refactoring.

Developers Are Already Shipping From Their Phones

The response from the developer community has been enthusiastic. Users like Lee Robinson (VP of Product at Vercel) and developer Ryo Lu shared demos of adding features and shipping code remotely — including from their phones. Engineers are reporting they can now ship dozens of pull requests per week with agent assistance.

One viral post summed up the reaction: "So you're telling me a VSCode clone can not only review my code but also test the feature on a cloud computer and send me a demo video of the whole process."

The Bigger Picture

Cursor's cloud agents represent a significant shift in how software development could work. Instead of writing every line of code yourself, you describe what needs to be done, and an AI agent handles the implementation, testing, and documentation. The developer's role shifts from writer to reviewer.

This raises obvious questions about the future of junior developer roles. If agents can handle routine coding tasks, test them, and even produce video demos, the entry-level programming job looks very different. The skill that matters becomes the ability to specify requirements clearly and review AI-generated code effectively.

The Competition Heats Up

Cursor isn't alone in this space. GitHub Copilot has been expanding its agent capabilities, and tools like Devin and Codex are pursuing similar autonomous coding approaches. But Cursor's video demo feature and deep IDE integration give it a unique edge — developers can see exactly what the agent built without reading through code changes.

Bottom Line

Cursor's autonomous cloud agents mark a new chapter in AI-assisted development. They write code, run tests, iterate on failures, and deliver pull requests with video proof of their work. With 30% of Cursor's own merges already agent-powered, this isn't a future promise — it's happening now. The question isn't whether AI agents will change software development, but how fast the shift happens and what it means for the developers who need to adapt.