Browser MCP Server

Chrome sessions already in use become the starting point for browser-mcp, rather than a separate headless browser. Through its extension, MCP clients can navigate pages, read text or HTML and capture screenshots. Form workflows have dedicated handling for autocomplete controls, custom dropdowns, date pickers and uploads. For mailbox views, transaction histories or billing tables that render rows only as they become visible, browser_extract_list scrolls the container and returns deduplicated row text with an end-of-list indicator. These tools support web application testing and work on platforms without APIs.

Authentication can involve retrieving an emailed code from an open Gmail tab, or requesting credentials and OAuth consent through a user-facing overlay. CAPTCHA tools attempt automated interaction, return screenshots for visual analysis and provide a human fallback. For application inspection, agents can watch for completed network requests, inspect console messages and read or change cookies and local storage. Provider token extraction navigates to API settings pages where tokens can be read. Concurrent sessions use color-coded groups to keep their tabs separate.