Skip to content
INTERNET INFRASTRUCTURE & NETWORKS

Cloudflare Open-Sources Kitesurf, a Stateless Browser Built Specifically for AI Agents

Cloudflare has officially fulfilled its ongoing promise to open-source Kitesurf, introducing the stateless browser engine to the developer community under the permissive Apache 2.0 license. Initially unveiled in August alongside a subsequent feature update in September that brought WebMCP support, improved compatibility, and new APIs, Kitesurf is designed from the ground up to run entirely on top of Cloudflare Workers. With this open-source release, developers now have the full ability to inspect the underlying source code, run the engine locally for testing and debugging, modify it to suit specific needs, and deploy custom instances directly to their own Cloudflare accounts.

The project marks a significant step forward in how automated systems and AI agents interact with the web. Traditional browser automation relies on heavy, resource-intensive containers running full browser suites like Chromium. In contrast, Kitesurf reimagines web interaction by executing core browser tasks natively within the serverless architecture of Cloudflare’s global network.

A Browser Built for Agents

Modern AI agents frequently require a browser interface to navigate the web, extract data, execute JavaScript, and interact with complex web applications. However, using full-fledged desktop browsers for these backend tasks introduces substantial performance hurdles. They consume massive amounts of CPU and memory, making it difficult and expensive to scale automation workloads concurrently.

Kitesurf was built to solve this exact problem by providing a lightweight, highly efficient alternative. Operating entirely on Cloudflare Workers, the browser engine parses HTML, applies CSS stylesheets, executes page-level JavaScript, and generates visual screenshots without ever needing a traditional browser container. Furthermore, Kitesurf natively supports the Chrome DevTools Protocol (CDP), allowing developers to seamlessly connect existing automation clients like Puppeteer or point sophisticated AI harness frameworks directly to the system.

By leveraging Cloudflare’s distributed infrastructure, Kitesurf scales dynamically on demand across the global network. This architecture allows developers to achieve rapid page loads and reliable automation while utilizing a fraction of the computational resources demanded by standard containerized browser instances.

What is Inside the Repository?

The newly published GitHub repository brings together Cloudflare’s extensive Developer Platform and APIs alongside a robust ecosystem of open-source Rust packages compiled to WebAssembly. These underlying technologies form the technical foundation of the browser, with all third-party components properly documented under the repository’s official NOTICE file to respect their original licenses.

Architecturally, Kitesurf is divided into three primary functional components. The Engine is responsible for managing active browser sessions and handling client connections. The PageScript component takes charge of running the target web page and executing its embedded JavaScript. Finally, the PageRenderer handles the visual output, drawing the final result.

The source code directory structure clearly separates these responsibilities. The packages folder houses individual components, including the core engine, the Rust-based foundation, custom DOM shims, the visual renderer, and the dedicated Kitesurf CLI. Developers can explore these directories to understand how the disparate pieces integrate into a cohesive serverless browser environment.

Kitesurf is now open source

Running Your Own Instance of Kitesurf

Developers interested in experimenting with the technology locally can do so by cloning the official GitHub repository and satisfying the documented build requirements. After pulling the source code, dependencies can be managed using standard package managers to initialize the development environment.

Building the project for the first time triggers the compilation of the Rust core and its associated dependencies, a process that requires a brief initial wait before caching subsequent builds for faster iteration. Once compiled, developers can use Wrangler to launch a local development server. This local instance exposes an endpoint that can be accessed directly via Chrome DevTools or integrated programmatically into existing automation scripts.

Integration with popular automation tools is straightforward. For instance, developers can utilize Puppeteer in a Node.js script by establishing a secure WebSocket connection to the local Kitesurf endpoint. Through this connection, scripts can spawn new pages, navigate to arbitrary URLs, and capture screenshots locally. Alternatively, engineering teams can point their AI agent harnesses directly to the local CDP port, granting automated agents full programmatic control over the Kitesurf browser engine.

Seamless Deployment to Cloudflare Accounts

Once local testing and modifications are complete, developers are not restricted to managed application programming interfaces. Kitesurf can be deployed directly to a personal Cloudflare account, allowing teams to run and scale their own instances of the stateless browser engine.

Because Kitesurf leverages Dynamic Workers to spin up isolated, secure worker environments for individual web pages, deploying the engine to a production account requires a Workers Paid plan. This architectural isolation ensures that different browser sessions remain secure and separate from one another while operating at the edge.

Community Involvement and Future Development

As Kitesurf enters its open-source phase, Cloudflare has emphasized that the project remains under heavy active development. Current engineering priorities center heavily on improving overall engine reliability, expanding web standards compatibility—as tracked through Web Platform Test metrics—and driving further efficiency gains.

The development team welcomes forks, bug reports, and pull requests from the wider developer community. Contributors are encouraged to consult the project’s formal guidelines before submitting code, with a recommendation to open preliminary issues featuring minimal reproductions or clearly defined scopes to streamline the review process. Contributions extend beyond writing code; developers can actively help by submitting simple web pages that reproduce rendering bugs, reporting failing automation workflows, or refining documentation.

As the project continues to evolve rapidly, the team plans to share ongoing technical updates through official channels, including company blogs, developer changelogs, and the dedicated Kitesurf landing page, where developers can monitor benchmark results and standards progress.