GitHub MCP Server

A Model Context Protocol (MCP) server that provides tools for interacting with GitHub's API. Currently supports creating repositories with descriptions, topics, and website URLs.

Features

Create GitHub repositories with auto-generated names from descriptions
Add topics/tags to repositories
Set repository homepages
Auto-initialize repositories with README files

Installation

Clone the repository
Install dependencies:

npm install

Build the server:

npm run build

Configuration

The server requires a GitHub personal access token with repository creation permissions. Add the following to your MCP settings file:

{ "mcpServers": { "github": { "command": "node", "args": ["path/to/github-server/build/index.js"], "env": { "GITHUB_TOKEN": "your-github-token" } } } }

Available Tools

create_repo

Create or update GitHub repositories using natural language commands.

Command Format

The tool accepts natural language commands for different operations:

Creating repositories:

Create a repository for [description] with tags [tag1 tag2 tag3] website [url]

or

Make a new repository called [description] tagged with [tag1, tag2, tag3]

Updating repository description:

Update [owner/repo] description to [new description]

or

Change [repo-name] description as [new description]

Updating repository tags:

Update [owner/repo] tags to [tag1 tag2 tag3]

or

Set [repo-name] topics as [tag1, tag2, tag3]

Updating repository website:

Update [owner/repo] website to [url]

or

Set [repo-name] homepage as [url]

Example Usage

Creating a new repository:

const result = await use_mcp_tool({ server_name: "github", tool_name: "create_repo", arguments: { command: "Create a repository for my machine learning image classifier with tags python tensorflow computer-vision website https://example.com/docs" } });

This will:

Create a repository named "my-machine-learning-image-classifier"
Set the description as "my machine learning image classifier"
Add "python", "tensorflow", and "computer-vision" as repository topics
Set the website to "https://example.com/docs"
Initialize with a README file

Updating repository description:

const result = await use_mcp_tool({ server_name: "github", tool_name: "create_repo", arguments: { command: "Update username/existing-repo description to Updated ML project for image classification" } });

Updating repository tags:

const result = await use_mcp_tool({ server_name: "github", tool_name: "create_repo", arguments: { command: "Update username/existing-repo tags to machine-learning python updated" } });

Updating repository website:

const result = await use_mcp_tool({ server_name: "github", tool_name: "create_repo", arguments: { command: "Update username/existing-repo website to https://example.com/new-docs" } });

The tool understands various natural language patterns and keywords: