What changed
Cloudflare has released Kitesurf, a stateless browser for AI agents, as open source on GitHub under the Apache 2.0 license. Developers can now run, modify, and deploy their own instance of the browser engine on Cloudflare Workers. The project was previously announced in August and further developed in September with WebMCP support, new APIs, and improved compatibility.
What this means in practice
Kitesurf runs entirely on Cloudflare Workers without a browser container. It parses HTML, applies CSS, executes JavaScript, and renders screenshots, supporting the Chrome DevTools Protocol (CDP) for client connections like Puppeteer. The browser is designed to be lightweight and efficient, using less CPU and memory than a full browser like Chromium, and scaling on demand across Cloudflare’s global Workers network.
Kitesurf is built using Cloudflare’s Developer Platform and APIs, along with open-source Rust packages compiled to WebAssembly. These include Blitz for HTML parsing, Stylo for CSS, Boa for JavaScript execution, Linebender for text layout, and Krilla for PDF output. The repository includes the Engine, PageScript, and PageRenderer components, with all packages and code organized in the repository’s packages directory.
Limitations and availability
To run Kitesurf locally, developers must follow the repository’s build requirements, clone the repository, install dependencies, and start the engine. A local development server is available at http://127.0.0.1:8787. For deployment, a Workers Paid plan is required due to the use of Dynamic Workers for isolated page execution. Developers can deploy the engine to their own Cloudflare account using the pnpm run deploy command.
Kitesurf is under active development, with a focus on improving compatibility, reliability, and efficiency. Contributions are welcome, but maintainers request focused fixes with tests, clear evidence, and well-defined scopes. The project is evolving quickly, with updates expected in the blog and changelog, and the official landing page for standards compatibility and benchmarks.
