Github Ranker Remote

A remote MCP server built with FastMCP that ranks GitHub candidates against a job's required skills.

What it does

Give it a list of GitHub usernames, a job title, and required skills — it fetches each profile from the GitHub API, scores them by skill match (languages + repo topics), and returns ranked candidate data. A prompt primitive is included for the MCP client's LLM to format the output as a markdown report.

Demo

Primitives

Primitive	Name	Description
@mcp.tool	rank_candidates	Fetches GitHub profiles, scores each against required skills, returns ranked JSON
@mcp.prompt	ranking_report	Takes the tool output and returns a prompt for the client LLM to generate a markdown report

Connecting

Go to Claude.ai → Settings → Customize → Add MCP Server → paste:

https://github-ranker-remote-mcp-server.onrender.com/mcp

Note: Hosted on Render's free tier — first request after 15 min idle takes ~30s to wake up.

Usage

Once connected, call the tool with:

• usernames — list of GitHub usernames
• job_title — role being hired for
• skills — comma-separated, using GitHub language names (Python, Go, TypeScript) and lowercase repo topic keywords (docker, fastapi, kubernetes)

Skills are matched against each user's top languages and repo topic tags.

Then pass the output into the ranking_report prompt — the client LLM generates the final markdown report. No external API call is made from the server.

Stack

• FastMCP — MCP server framework
• httpx — async GitHub API calls
• Python 3.11+
• Deployed on Render

Running locally

uv init
uv add fastmcp httpx

Create a .env file:

GITHUB_TOKEN=your_token_here

uv run python server.py

Server starts on http://0.0.0.0:8000/mcp.

Development / testing

uv run fastmcp dev inspector --server-spec server.py

Opens the MCP Inspector UI for interactive tool testing.