MCP.so
ログイン
M

mcp-test

@w-jeon

mcp-test について

mcp test

基本情報

カテゴリ

開発者ツール

トランスポート

stdio

公開者

w-jeon

投稿者

Wendy Jeon

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "mcp-framework-w-jeon": {
      "command": "docker",
      "args": [
        "compose",
        "up",
        "--build",
        "-d"
      ]
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is mcp-test?

mcp-test is a framework for creating custom MCP tools that interact with large language models. It integrates file processing (PDF, Word, Excel) and web content fetching, designed for extending tools like Cursor IDE.

How to use mcp-test?

Deploy via Docker Compose or traditional Python. After setup, invoke tools by calling their names: file, pdf, word, excel, or url with the required parameters such as a file path or URL.

Key features of mcp-test

  • Automatic file type identification and handling
  • PDF processing with quick and full modes
  • Word document text, table, and image extraction
  • Excel multi-sheet parsing with structure info
  • Web page content retrieval with HTTP error handling
  • Modular tool architecture for easy extension

Use cases of mcp-test

  • Extract text and images from PDF documents for LLM analysis
  • Parse Word files to retrieve paragraphs, tables, and embedded images
  • Convert Excel spreadsheets into structured data for automated workflows
  • Fetch web page content as plain text for summarization or data collection
  • Integrate as an MCP server in Cursor IDE to add custom document processing

FAQ from mcp-test

How does mcp-test handle PDFs compared to other tools?

It uses PyMuPDF as the primary engine with fallback to pymupdf4llm and PyPDF2. It offers two modes: quick (text only, up to 50 pages) and full (text + images, up to 30 pages).

What are the runtime dependencies?

Python 3, plus packages: mcp, PyMuPDF, python-docx, pandas, openpyxl, httpx, anyio, click. System libraries like poppler and tesseract may be needed for advanced OCR.

Where does processed data live?

All processing happens locally on the server. Temporary files are created for large documents and automatically cleaned up.

Are there any known limits?

PDF full mode is capped at 30 pages, quick mode at 50 pages. HTTP requests have a 5‑minute timeout. Images in PDFs are downscaled to a maximum DPI and size.

What transport does mcp-test support?

It exposes an SSE (Server‑Sent Events) endpoint, typically at http://localhost:8000/sse. No authentication is mentioned in the provided documentation.

コメント

「開発者ツール」の他のコンテンツ