neverthesameagain/Personalised-MCP-server

Personal MCP Server for AI Resume Submission

This repository contains a Model Context Protocol (MCP) server used to securely deliver my resume to Puch AI for application validation and submission. It doubles as a neat personal project demonstrating token-based auth, LLM-ready content pipelines, and cloud deployment.

Features

• Bearer Token Auth – Simple provider that validates a single application key.
• Resume Tool – Reads a local PDF/DOCX/TXT resume, converts it to Markdown, and returns raw MD text (no extra formatting).
• Fetch Tool – Retrieves webpages and simplifies HTML to Markdown for LLMs; supports pagination of long pages.
• Cloud Ready – Designed for one-click deploy on Render; works locally with uvicorn.

Tech Stack

• Python 3.11+
• fastmcp (MCP server framework)
• httpx, readabilipy, markdownify (content fetching & cleaning)
• pypandoc (document → Markdown conversion)
• Render (hosting)

Project Structure

.
├── mcp_server.py       # Main server
├── requirements.txt    # Python dependencies
├── Procfile            # Start command for Render
└── README.md           # This file

⚙Configuration

The server expects three values in the source file. For convenience, they are defined as constants near the top of mcp_server.py.

• TOKEN – Your Puch application key (from /apply <URL>)
• MY_NUMBER – Phone number in E.164 without the plus (e.g., 917988236305)
• resume_path – Absolute path to your resume (PDF/DOCX/TXT)

Security tip: You may move TOKEN/MY_NUMBER to environment variables instead of hardcoding.

Example (env-based):

export TOKEN="<your_key>"
export MY_NUMBER="<countrycode><number>"  # e.g. 917988236305

Then in mcp_server.py, read them via os.getenv("TOKEN") and os.getenv("MY_NUMBER") with sensible fallbacks.

Quick Start (Local)

1. Install system Pandoc (required by pypandoc):

   • macOS: brew install pandoc
   • Ubuntu/Debian: sudo apt-get update && sudo apt-get install -y pandoc
   • Windows (choco): choco install pandoc

2. Install Python dependencies

   pip install -r requirements.txt
   

3. Edit mcp_server.py

   • Set TOKEN, MY_NUMBER, and resume_path.

4. Run

   python mcp_server.py
   

   You should see Uvicorn running on http://0.0.0.0:8085.

5. (Optional) Tunnel for public URL If testing locally, you can use ngrok:

   ngrok http 8085
   

☁Deploy on Render (Recommended)

1. Push this repo to GitHub.

2. Go to Render → New → Web Service.

3. Connect the repository.

4. Configure:

   • Runtime: Python 3.11+
   • Build Command: pip install -r requirements.txt
   • Start Command: python mcp_server.py
   • Instance: Free

5. (Optional) Add Environment Variables:

   • TOKEN = your application key
   • MY_NUMBER = e.g., 917988236305
   • RESUME_PATH = absolute path in the container (if you commit your resume or mount storage)

6. Deploy. Render will output a URL like:

   https://your-service.onrender.com
   

Note: If your resume is private and not in the repo, consider Render Persistent Disks or upload during build via private storage. For the application step, committing a temporary PDF to the repo is simplest; remove it after submission.

MCP Tools / Endpoints

This server is MCP-compliant and exposes tools (consumed via /mcp by Puch):

validate

• Description: Returns the configured phone number.
• Purpose: Puch uses it to validate {TOKEN} and phone pairing.

resume

• Description: Returns the resume as Markdown string.
• Input: none (reads the file from resume_path)
• Output: Pure Markdown text (no extra formatting or wrapping).

fetch

• Description: Fetch a URL, simplify HTML → Markdown, paginate long outputs.

• Args:

  • url (string, required)
  • max_length (int, default 5000)
  • start_index (int, default 0)
  • raw (bool, default False) – if True, returns raw HTML

Connect from Puch

Once the service is public:

/mcp connect https://your-service.onrender.com/mcp <YOUR_TOKEN>

Quick Smoke Test (Without Puch)

You can sanity-check that the service is up (root will 404, that’s expected). For real checks, connect via Puch or hit the MCP route with an authorized client. If you must curl, confirm the service port and health by requesting / and expecting 404 rather than connection refused.

Troubleshooting

• 404 Not Found at / – Expected. Use /mcp with an MCP client (e.g., Puch).
• Failed to read or convert resume – Ensure Pandoc is installed and resume_path is correct. Try converting locally with pandoc <file> -t gfm -o /tmp/test.md.
• Render build fails – Check Python version, requirements.txt, and that pandoc is not required at build time (only at runtime via pypandoc).
• Auth fails in Puch – Verify you passed the correct token and that validate returns the exact phone format {countrycode}{number} with no plus sign.

Security Notes

• The included SimpleBearerAuthProvider validates a single bearer token. For production, use a proper key management solution.
• Avoid committing your real token or phone number. Prefer environment variables.
• Remove your resume from the repository once the application process is complete.