Git MCP Go

Git Remote MCP is a high-performance Model Context Protocol (MCP) server written in Go. It empowers AI Agents (Claude Desktop, Gemini CLI, Cursor, Qwen) to explore, search, and analyze GitHub repositories in real-time without the need for local cloning.

The Problem It Solves

AI models often suffer from a "knowledge cutoff" or hallucinate library versions. This tool provides Real-Time Context:

Outdated Dependencies: AI can fetch the absolute latest SemVer-sorted tags to recommend up-to-date libraries.
Context Gap: AI can "read" remote source code, file structures, and documentation to understand libraries it wasn't trained on.
Blind Coding: Instead of guessing APIs, the AI can search and read the actual implementation in the repository.

Why Go?

Zero System Dependencies: No OpenSSL or C-bindings required (unlike reqwest in Rust which can have cross-compilation issues).
Super Fast Compilation: Rebuilds complete in under 0.5 seconds.
Easy Cross-Compilation: Build for any platform with simple GOOS/GOARCH environment variables.
Concurrency Ready: Future features can leverage Go's goroutines for parallel repository operations.

Features

Zero-Clone Exploration: Fetch trees, files, and metadata via GitHub API instantly.
Smart Dependency Solving: Fetches and sorts tags by Semantic Versioning (SemVer), ensuring the AI suggests the actual latest version.
Semantic Search: Find functions, structs, or text across the entire repository using GitHub's Search API.
Secure & Scalable: Natively supports GITHUB_TOKEN authentication to increase API rate limits from 60 to 5,000 requests/hour.
Multi-Arch Support: Cross-compile for x86_64, aarch64 (ARM64), and armv7 easily.

Tools Available for AI

| Tool | Description | |------|-------------| | get_tags | Returns latest tags/versions. Supports limit and SemVer sorting (e.g., v1.10 > v1.9). | | search_repository | Search for code, specific functions, or text definitions within the repo. | | get_file_tree | Recursively lists files to reveal project architecture/structure. | | get_file_content | Reads the raw content of specific files from any branch/tag. | | get_readme | Automatically fetches the default README for a quick project overview. | | get_changelog | Compares two tags and returns a summary of commit messages. |

Visualizing the Data Flow

sequenceDiagram
    participant User
    participant AI as AI Agent
    participant MCP as Git-Remote-MCP
    participant GitHub as GitHub API

    User->>AI: "Use library X in my project"
    AI->>MCP: get_tags(url_X, limit=5)
    MCP->>GitHub: Request Tags Metadata (Auth Token)
    GitHub-->>MCP: Raw Tags Data
    Note over MCP: Sorting via SemVer Logic
    MCP-->>AI: ["v2.1.0", "v2.0.9", ...]

    AI->>MCP: search_repository(query="struct Config")
    MCP->>GitHub: Search Code API
    GitHub-->>MCP: Result Paths
    MCP-->>AI: Found in "src/config.rs"

    AI->>User: "Here is the code using v2.1.0 and correct config struct"

Installation

Option A: Quick Install (Binary)

Install the pre-compiled binary for your OS and Architecture (Linux x86_64/aarch64/armv7 or macOS Intel/Silicon):

curl -fsSL https://raw.githubusercontent.com/HanSoBored/git-mcp-go/main/installation/install.sh | bash

Option B: Build from Source (Go)

Requires Go 1.21+:

git clone https://github.com/HanSoBored/git-mcp-go.git
cd git-mcp-go
./installation/build.sh

Option C: Manual Build

cd src/
go mod download
go build -ldflags="-s -w" -o git_mcp

# Cross-compile for Linux x86_64
GOOS=linux GOARCH=amd64 go build -ldflags="-s -w" -o git_mcp_linux_amd64

# Cross-compile for macOS ARM64
GOOS=darwin GOARCH=arm64 go build -ldflags="-s -w" -o git_mcp_darwin_arm64

Configuration

Add this to your MCP client configuration (Claude Desktop, Cursor, Gemini CLI, etc.):

{
  "mcpServers": {
    "git-remote": {
      "command": "/usr/local/bin/git_mcp",
      "args": [],
      "env": {
        "GITHUB_TOKEN": "ghp_your_personal_access_token_here"
      }
    }
  }
}

Important: Adding a GITHUB_TOKEN is highly recommended to avoid the 60 requests/hour rate limit. With authentication, you get 5,000 requests/hour.

System Prompt for AI Agents

To maximize the utility of this MCP, add this to your Agent's system instructions:

You are equipped with the 'git-remote' MCP toolset.
1. When asked about a library/dependency, ALWAYS use 'get_tags' with 'limit: 5' to verify the latest version. Do not guess.
2. Before suggesting implementation details, use 'search_repository' to find relevant code definitions (structs, functions).
3. Use 'get_file_content' to read the actual code context before answering.
4. If a user asks about a project structure, use 'get_file_tree'.

Example Usage

Check Latest Version

User: "What's the latest version of clap?"
AI: [Calls get_tags with url="https://github.com/clap-rs/clap", limit=5]
AI: "The latest version is v4.5.59"

Search for Code

User: "How does this project handle dlopen?"
AI: [Calls search_repository with query="dlopen"]
AI: [Calls get_file_content to read the relevant file]
AI: "Here's how dlopen is handled..."

License

MIT License. Feel free to use and contribute!

MCP Servers