mozilla/firefox-devtools-mcp

What it solves

This project allows AI agents to automate and interact with the Firefox browser. It bridges the gap between LLM-based clients (like Claude Code or Cursor) and the browser's internal developer tools, enabling agents to perform web navigation, testing, and debugging tasks automatically.

How it works

It implements a Model Context Protocol (MCP) server that communicates with Firefox via WebDriver BiDi and Selenium WebDriver. The server exposes a set of tools to the AI agent, allowing it to control the browser, inspect network requests, take screenshots, and execute scripts. It can run in headless mode, attach to an existing browser session, or even automate Firefox on Android devices via ADB.

Who it’s for

• AI Developers: Those building agents that need to browse the web or test web applications.
• QA Engineers: Users wanting to automate browser-based testing using natural language agents.
• Web Developers: People who want to use AI to debug network requests, console logs, or page performance.

Highlights

• Comprehensive Toolset: Includes tools for page navigation, element interaction (click/fill), network monitoring, console log retrieval, and screencasting.
• Flexible Presets: Offers different tool levels (slim, basic, developer, mozilla) to balance functionality and security.
• Context Management: Supports saving bulky output (like screenshots or large logs) directly to disk to avoid exhausting the LLM's context window.
• Cross-Platform Automation: Supports desktop Firefox and Firefox for Android.
• Existing Session Support: Can connect to a running Firefox instance to maintain cookies and active logins.