OpenBrowser-AI: Direct Browser Control for AI Agents via Chrome DevTools Protocol
By
Billy Enrizky
There's a fresh bagel in here somewhere. We couldn't find it.
Summary
OpenBrowser-AI is a tool that connects AI agents directly to web browsers using raw Chrome DevTools Protocol (CDP) without abstraction layers. It enables LLMs to write Python code in a persistent namespace, batching operations per call while maintaining page state in about 450 characters. The tool has been benchmarked against three other frameworks on six real tasks, achieving 100% accuracy, using 2.6x fewer tokens, and reducing inference costs by 59%. It's MIT licensed, includes CLI and MCP server support, works with 15 LLM providers, and has published reinforcement learning studies for training open-source browser control models.
Key quotes
· 5 pulledOpenBrowser connects AI agents to browser through raw CDP. No abstraction layer.
The LLM writes Python in a persistent namespace, batching operations per call. Page state at ~450 characters.
Benchmarked against 3 frameworks on 6 real tasks: 100% accuracy across the board, 2.6x fewer tokens, 59% lower inference costs.
Methodology is public and reproducible. MIT licensed. CLI + MCP server. 15 LLM providers.
Two published RL studies training open-source models for browser control.
You might also wanna read
BrowserOS: Open-Source Chromium Browser with Native AI Agent Support
BrowserOS is an open-source Chromium-based browser that runs AI agents natively, positioning itself as a privacy-first alternative to commer
github.com·4mo ago
Agent Browser Protocol (ABP): Deterministic Browser Automation for AI Agents
The article introduces ABP (Agent Browser Protocol), a Chromium fork designed for deterministic browser automation that works seamlessly wit
github.com·2mo ago
Browser Harness: A lightweight CDP-based tool for LLM-driven browser automation
Browser Harness is a lightweight, self-healing browser automation tool built on Chrome DevTools Protocol (CDP) that gives LLMs direct contro
github.com·1mo ago
Why Browser Development Has Become a Benchmark Test for AI Systems
The article discusses why people are suddenly building browsers with AI, explaining that browser development serves as an ideal test case fo
news.ycombinator.com·4mo ago
Overcoming Barriers to Open-Source Agents with Browser-Runnable AI Agents
The article discusses overcoming barriers to open-source agents by enabling them to run in browsers as HTML files.
blog.mozilla.ai·11mo ago
Web Browsers as Secure Sandboxes for AI Coding Agents
The article discusses how web browsers serve as an ideal sandbox environment for AI coding agents, leveraging 30 years of development in run