Show HN: Deterministic browser control for AI agents (~90% on Mind2Web)

theredsix

Hi HN, op here! This is an open source browser protocol for LLM agents. The browser shows the model the current page, the model chooses the next action, and the browser returns the new state. Between steps, JavaScript and time are frozen so the page stays still while the model thinks. That makes things like ecommerce shopping and popup-heavy web app workflows much more reliable. Using this setup, the project gets ~90% on Online Mind2Web. My bet is that browser agents need a protocol designed for models, not just wrappers around CDP.

bignoggins

what do you do differently compared to other options?

shane-moran

This is a great example of breaking the trend thinking it’s always model improvement needed, but sometimes the model doesn’t have the best way to interface with the data or system. The cleanliness of this approach that improves the ability for the model to interact without having to completely redefine the interface system with respect to still being able to use websites and the computer as is and not having to develop an entirely new interface protocol at the machine level.

Thors3n

Very exciting stuff! Most agent browser stacks still feel clunky to me. This is very promising, turning browsing into deterministic, atomic steps should definitely improve user interaction and E2E utility.

greggberry

This project is incredible! I already have it set up for local Claude agent use and seeing significant improvement, both in accuracy and task efficiency: `claude mcp add browser -- npx -y agent-browser-protocol@rc --mcp` Additionally, if you want to configure with Claude Desktop, add the following to your `claude_desktop_config.json` after installing the MCP: ``` "mcpServers": { "browser": { "command": "npx", "args": [ "-y", "agent-browser-protocol@rc", "--mcp" ] } } ```