Tender Documents Handler MCP (FileResource-compatible)

This is a FastMCP v2 server designed for FastMCP Cloud and Tri-Tender.

Key characteristics:

• ✅ Uses FileResource tool parameters – no local file paths
• ✅ Returns LLM-ready strings (not JSON) for maximum compatibility
• ✅ Lightweight dependencies (no pymupdf, no heavy native PDF engines)
• ✅ Safe imports on FastMCP Cloud (no import-time crashes)

Entrypoint (FastMCP Cloud)

Use this as your entrypoint:

server.py:mcp

Tools

parse_document(file: FileResource) -> str

Extract raw text from an uploaded document. Accepts:

• PDF
• DOCX
• TXT / HTML (as text)
• Images (basic OCR if Tesseract is available)

Returns plain text.

analyze_tender(file: FileResource) -> str

Main Tri-Tender analysis tool.

• Reads the file bytes
• Detects mime
• Extracts text
• Adds a system-style instruction block telling the LLM how to interpret the tender
• Appends the full tender text

The result is ideal to drop directly into the model as context.

extract_metadata(file: FileResource) -> str

Lightweight metadata: mime, size, page count (for PDFs) and usage instructions.

validate_document(file: FileResource) -> str

Quick pre-flight sanity check – verifies that text can be extracted and returns a small preview snippet.

Resources

• tender://config/version
• tender://config/supported-formats

Running Locally

uv pip install fastmcp
pip install -r requirements.txt

fastmcp run server.py

Or:

python server.py