GitHub

You run a Minimal Model Context Protocol (MCP) server that exposes GitHub-focused tools for AI assistants. It lets you perform automated code reviews, PR analysis, and repository insights through a streamlined, MCP-friendly interface. This guide shows you how to install, run, and connect to the server from your MCP client.

How to use

You connect to the server from an MCP client and provide credentials to authorize GitHub access. Run the server locally, then point your client to the local MCP endpoint. The default listening port is 3000 unless your client specifies otherwise.

How to install

Prerequisites: you need Node.js and a package manager installed on your system. The server is started after dependencies are installed and the credentials are configured.

# 1. Clone & install
 git clone <repo-url>
 cd github-review-mcp
 pnpm install

# 2. Add credentials
 echo "GITHUB_TOKEN=ghp_your_token_here" > .env

# 3. Run the server
 pnpm start

Additional notes

The server listens on the port specified by your MCP client (default 3000). You can also run the server in a container and expose port 3000 to access it from your MCP client. The MCP client configuration demonstrates how to launch the server as a local stdio process using the node runtime.

# MCP client snippet (stdio)
{
  "mcpServers": {
    "github_review": {
      "command": "node",
      "args": ["/absolute/path/github-review-mcp/src/index.js"],
      "env": { "GITHUB_TOKEN": "ghp_your_token_here" }
    }
  }
}

Docker (optional)

If you prefer running the server in a container, you can build and run a image locally. The container should receive your GitHub token and expose port 3000 so your MCP client can connect.

# Build
 docker build -t gh-mcp .

# Run
 docker run -e GITHUB_TOKEN=ghp_your_token_here -p 3000:3000 gh-mcp

Environment variables

Configure these values to tailor behavior and resilience of the MCP server. Most variables are optional, with sensible defaults when not provided.

GITHUB_TOKEN        - GitHub Personal Access Token (required)
MAX_PATCH_SIZE      - Maximum diff patch size (chars) (default 2000)
MAX_FILES_TO_REVIEW - Maximum files processed per PR (default 50)
REQUEST_TIMEOUT     - HTTP request timeout (ms) (default 30000)
LOG_LEVEL           - Logging level (debug, info, …) (default info)
ENABLE_DEBUG        - Verbose logging (true/false) (default false)

Security and best practices

Keep your GitHub token secure. Do not commit credentials into your source control. Prefer environment-based configuration and rotate tokens regularly. If you deploy via Docker or other orchestration, ensure token values are provided through secure secrets management.

Usage from an MCP client

From an MCP client, point to the local or remote MCP endpoint and provide the required credentials. You can either run the server as a local process with node or connect to a hosted MCP endpoint if available.

Available tools

get_review_prompts

Primary entry point to generate review prompts used for guiding PR analysis.

get_pr_details

Fetch detailed information about a pull request, including author, title, and status.

get_pr_files

Retrieve the list and metadata of files changed in a PR.

get_pr_commits

Obtain the commit history associated with a PR.

get_file_content

Read the content of a specific file at a given path and revision.

post_pr_review

Submit a review back to the PR with comments or status updates.

get_repo_info

Query repository-level information such as owners and visibility.

analyze_code_quality

Analyze code quality and style patterns across changed files.

analyze_diff_impact

Assess the potential impact of diffs on code paths and behavior.

detect_security_issues

Identify potential security vulnerabilities in the changed code.

detect_code_patterns

Spot common code patterns that may indicate anti-patterns or risks.

analyze_dependencies

Review dependencies for freshness, licensing, and risk.

analyze_test_coverage

Evaluate test coverage metrics for changed components.

generate_suggestions

Provide actionable improvement suggestions based on analysis.