anomalyco/browse
MCP SERVER FOR BROWSER CONTROL BY AI AGENTS
MCP server and Chromium extension allowing AI agents to control your real browser: open tabs, click, type, take screenshots, and hand back control for login or 2FA.
// KEY FEATURES
- Agents open background tabs, read pages, click and type text
- Screenshots, network traffic capture and screen recording support
- Tab handed to user for login, 2FA or payment with Continue button
- Works in Chrome, Edge, Brave, Arc, Vivaldi, Opera and Chromium
// DEEP DIVE: HOW IT WORKS
The project is a native MCP server `browse` and a Chromium extension. Agents control the browser via MCP: open background tabs, read, click, type, take screenshots, capture network traffic and record screen. When login, 2FA or payment is needed, the tab is handed to the user, and after clicking Continue the agent continues.
It uses your real browser with profile, logins and extensions, not headless. Supports OpenCode, Claude Code, Codex, Cursor and any MCP client. Installation via single script or package managers: configures agent config, installs binary, pairs extension via Chrome Web Store or code. Binary auto-updates.
Requires Chromium-based browser (Chrome, Edge, Brave, Arc, Vivaldi, Opera). For npm — Node.js 22+. Extension from Chrome Web Store; without GUI — pairing code. Agent must support MCP; after install — agent restart or MCP reload.
// USE CASES
Best for: For developers integrating AI agents with web interfaces via MCP
- Automate data collection from sites requiring authentication via AI agent
- Test web applications by delegating clicks and text input to agent with your browser
- Conduct security audits by recording agent actions and network traffic in real browser
// QUICK START
curl -fsSL https://opencode.ai/browse/install | bashCommands are taken from the repository README — check the current version on GitHub before running.
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram