Browser for AI Agent

Let AI agents read browser tab content, control the browser (tabs, windows, and more), and call tab tools.

Control your browser from any agent Any agent that speaks MCP (or runs shell commands via a Skill) can: - Read page content, cookies, localStorage, page errors, screenshots, and more - Manage tabs and windows from a background script it writes itself - Run scripts in a tab and call page tools for automation - Chromium only: debug with CDP tools for network requests/responses and other low-level visibility When Codex / Claude Code / VS Code / Cursor / Zed are detected, the welcome page offers to configure MCP or install a Skill automatically. Page tools Agents can call tools scoped to the current tab: - Subscribed toolsets from the built-in marketplace (or paste any URL in settings), filtered by tab URL - Tools registered by page authors via the WebMCP API Security notes - Make sure your AI agent environment is not vulnerable to prompt injection — otherwise an attacker may read your browser data through the extension. - Only install page tools from sources you trust; a malicious tool can execute arbitrary scripts in your browser. Privacy Browser data is sent only to the local Native Messaging Host and your configured MCP client / AI agent. We do not sell user data or use it for advertising or unrelated purposes. Only connect AI agents you trust, and only install toolsets you trust.