camofox

CamoFox MCP acts as a bridge between your AI agent and the CamoFox Browser Server, providing an MCP REST-style interface that lets your agent open tabs, navigate pages, extract data, and manage sessions with anti-detection features. The MCP server wraps the CamoFox Browser Server and exposes commands like create_tab, navigate_and_snapshot, list_profiles, and close_tab, enabling your agent to perform web automation while preserving human-like behavior. To use it, start the CamoFox browser server (via Docker, npx, or from source) and run the MCP client configured to connect to the browser URL (default http://localhost:9377). Then point your MCP client at the camofox-mcp instance, supply the CAMOFOX_URL environment variable, and begin sending MCP commands through your editor or agent interface. Typical workflows include verifying connectivity with server_status, creating a tab for a target URL, navigating and capturing content, and managing user session data like cookies and profiles. The integration supports features such as search engine macros, cookie import/export, CSS selector fallbacks, YouTube transcript extraction, and snapshot pagination for long pages.

Prerequisites:

• Node.js 18+ (recommended) or Docker for a zero-config setup
• Access to a running CamoFox Browser Server (via Docker, npx, or source build)
• An MCP-compatible client (VS Code Copilot, Claude Desktop, Cursor, or other MCP clients)

Option 1 — Docker (recommended for ease):

1. Start the CamoFox Browser Server (running elsewhere): docker run -d -p 9377:9377 --name camofox-browser ghcr.io/redf0x1/camofox-browser:latest
2. Run the MCP server container: docker run -i --rm -e CAMOFOX_URL=http://localhost:9377 ghcr.io/redf0x1/camofox-mcp:latest

Option 2 — npx (quick start, Node.js 18+ required):

1. Start CamoFox Browser (in a terminal): npx camofox-browser@latest
2. Start the MCP server (in another terminal or background): npx camofox-mcp@latest
3. Verify the browser is reachable: curl http://localhost:9377/health

Option 3 — From Source (developers):

1. Clone and run CamoFox Browser: git clone https://github.com/redf0x1/camofox-browser.git cd camofox-browser && npm install && npm run build && npm start
2. Clone and build CamoFox MCP: git clone https://github.com/redf0x1/camofox-mcp.git cd camofox-mcp && npm install && npm run build
3. Ensure CAMOFOX_URL is set (e.g., http://localhost:9377) and start your MCP client configured to connect to the MCP server.

Configuring the MCP client (examples):

• VS Code (Copilot): create .vscode/mcp.json with command: "npx", args: ["-y", "camofox-mcp@latest"], env: {"CAMOFOX_URL": "http://localhost:9377"}
• Claude Desktop: use mcpServers with command: "npx", args: ["-y", "camofox-mcp@latest"], env: {"CAMOFOX_URL": "http://localhost:9377"}
• Cursor: similar to Claude, in ~/.cursor/mcp.json

Verification prompt (example to paste into your agent): Verify my CamoFox MCP setup. Run these checks and report results:

1. Call server_status — is the browser server connected?
2. If connected: create_tab with url https://example.com
3. navigate_and_snapshot on that tab (wait for text: "Example Domain")
4. list_profiles to confirm profile storage is accessible
5. close_tab for the test tab

Tips and common issues:

• Ensure CAMOFOX_URL is reachable from the MCP client; use 127.0.0.1 or hostnames as appropriate in your environment.
• When using Docker, make sure the browser server and MCP container can communicate on the same network or expose the correct host/ports.
• If server_status reports disconnected, restart the CamoFox Browser Server and then retry; verify that curl http://localhost:9377/health returns ok: true and browserConnected: true.
• For long-running automation, enable session persistence by importing cookies or maintaining login state across interactions as supported by the MCP server.
• Use the YouTube transcript and accessibility snapshot features to reduce token usage and improve data extraction efficiency.