drissionpage-mcp-browser-automation

Install the drissionpage-mcp package from PyPI and register it as an MCP server in your client configuration. The agent then calls tools like page_navigate, element_click, and element_type to control the browser deterministically.

Navigate with page_navigate, then use wait_for_element or wait_until with the stable condition for dynamic content to load. Extract results with element_find_all, which returns bounded lists with recommended selectors for each item.

Yes, it works with Claude Desktop, Cursor, and Codex by adding a drissionpage entry to the MCP settings JSON. For GUI-launched clients that lack shell PATH, specify the absolute Python path with the drissionpage_mcp.cli module.

Yes, it provides frame_list, frame_snapshot, and frame_find for iframe content, plus shadow_find and shadow_find_all for shadow DOM. These tools locate elements inside nested contexts without changing the main frame state.

Run drissionpage-mcp doctor to diagnose startup issues. Common causes include a missing Chrome executable, which you can fix by setting the CHROME_PATH environment variable to your Chrome or Chromium binary.

File uploads through element_upload_file are restricted to files under the DP_MCP_UPLOAD_ROOT directory as a security boundary. Set this environment variable before starting the server and place upload files inside that directory.