mineru_mcp_project

This MCP server wraps Mineru's PDF parsing capability into a FastMCP tool named convert_pdfs_with_mineru. It exposes a workflow that scans a folder for PDF files, uploads them to Mineru for OCR and conversion, and then downloads the resulting ZIP containing HTML and image assets. The server supports configurable language, table recognition, extra output formats, and optional renaming of image assets to keep HTML references consistent. To use it, start the service locally and connect with any MCP-compatible client. The convert_pdfs_with_mineru tool accepts parameters like pdf_folder, output_folder, api_token (or rely on the MINERU_API_TOKEN environment variable), language, enable_table, extra_formats, and more. After the job completes, you’ll receive a structured summary including upload status, playback of results, and the location of the output HTML and assets.

Prerequisites:\n- Python 3.10 or newer\n- Internet access to Mineru (mineru.net) and the ability to reach Mineru APIs\n\n1) Install the package in editable mode (from the repo root):\nbash\npip install -e .\n\n2) Provide Mineru API Token (either via env var or per-call):\nbash\n# Option A: set environment variable for the session\nexport MINERU_API_TOKEN=your_token_here\n# Option B: pass api_token in the call to convert_pdfs_with_mineru\n\n3) Start the MCP server:\nbash\n# If installed with the console script, you can use the wrapper:\nrun-mineru-mcp\n# Alternatively, run as a Python module (equivalent):\npython -m mineru_mcp_project\n\nThe service will listen by default at http://127.0.0.1:4399/mcp/ .

Environment variables and configuration:\n- MINERU_API_TOKEN: Mineru API token. Can be provided via this env var or per-call via api_token.\n- Ensure network access to https://mineru.net.\n\nCommon issues and tips:\n- If upload or conversion fails, verify the API token and network connectivity.\n- For long-running tasks, adjust max_wait and poll_interval accordingly in the convert_pdfs_with_mineru parameters.\n- If assets are not renamed in the output HTML, enable rename_assets_flag in the tool parameters and ensure the ZIP contains figure/images folders.\n- The server exposes convert_pdfs_with_mineru only after the service starts; ensure the MCP client calls the correct tool name and parameter schema.