Cloudflare open-sources Kitesurf, its stateless AI browser for Workers

A lighter alternative to heavy browser engines

Software development for artificial intelligence requires agents to read and interact with the web. Traditionally, running web automation for AI meant spinning up full headless instances of browsers like Chromium. These browser instances consume heavy amounts of CPU power and memory. They are often difficult to scale efficiently across distributed server networks.

Cloudflare initially announced Kitesurf to solve this issue. Kitesurf is a stateless browser engine designed specifically for AI agents. Instead of running inside a traditional heavy container, it runs directly on Cloudflare Workers. It can parse HTML code, apply CSS styling, run page JavaScript, and produce screenshot images. Because it supports the Chrome DevTools Protocol, existing automation tools like Puppeteer can connect to it directly.

Source code available under Apache 2.0

Cloudflare has now fulfilled its promise to release Kitesurf to the public. The source code is available on GitHub under the open-source Apache 2.0 license. Developers can now read the code, modify it, test it locally, or host their own instances on their personal Cloudflare accounts rather than relying entirely on managed platform APIs.

The underlying architecture of Kitesurf combines Cloudflare’s Developer Platform with several open-source Rust packages compiled into WebAssembly. It uses Blitz to parse HTML code and layout pages, Stylo for handling CSS cascades and computed styles, and Boa to run page JavaScript inside WebAssembly when isolated runtime environments are unavailable. Additional graphics and text layout libraries handle 2D visual elements and PDF generation.

Building and hosting custom instances

The codebase is structured into three main components. An engine manages overall sessions and network connections, PageScript executes page logic, and PageRenderer draws visual output. Developers can build the repository locally using standard JavaScript package tools and Cloudflare’s local development environment server.

For production deployments, users can push their modified engine directly to their Cloudflare accounts. Deploying Kitesurf requires a paid Workers account, as the architecture relies on Cloudflare’s Dynamic Workers feature to generate isolated environments for individual web pages. Cloudflare noted that Kitesurf remains under heavy development, with ongoing work focused on expanding Web Platform Test compatibility, improving rendering accuracy, and raising performance efficiency.

What it means for developers and platforms

For builders constructing AI platforms, an open-source lightweight browser reduces infrastructure reliance on traditional cloud server instances. Developers can run browser tasks closer to users without the overhead of heavy virtual machines or browser containers. Having access to the open-source codebase gives teams control over how their AI web scrapers and automation agents handle complex web pages.

For the broader cloud market, the release highlights a shift toward WebAssembly and serverless environments replacing standard server workloads. Investors and engineering leaders should monitor how quickly Kitesurf improves its web standard compatibility tests compared to traditional browser engines. Watching developer contributions on GitHub and integration updates on Discord will signal how rapidly lightweight browser runtimes could replace heavy Chromium containers for AI workflows.