Git Repo Research MCP Server

⚠️ DEPRECATION NOTICE: This server is deprecated and will no longer receive updates. For library documentation and code research, we recommend Context7 which provides up-to-date docs without requiring AWS credentials. See the migration guide for details.

Model Context Protocol (MCP) server for researching Git repositories using semantic search

This MCP server enables developers to research external Git repositories and influence their code generation without having to clone repositories to local projects. It provides tools to index, search, and explore Git repositories using semantic search powered by Amazon Bedrock and FAISS.

Features

• Repository Indexing: Create searchable FAISS indexes from local or remote Git repositories
• Semantic Search: Query repository content using natural language and retrieve relevant code snippets
• Repository Summary: Get directory structures and identify key files like READMEs
• GitHub Repository Search: Find repositories in AWS-related organizations filtered by licenses and keywords
• File Access: Access repository files and directories with support for both text and binary content

Prerequisites

Installation Requirements

1. Install uv from Astral or the GitHub README
2. Install Python 3.12 or newer using uv python install 3.12
3. • uv - Fast Python package installer and resolver
4. AWS credentials configured with Bedrock access
5. Node.js (for UVX installation support)

AWS Requirements

1. AWS CLI Configuration: You must have the AWS CLI configured with credentials that have access to Amazon Bedrock
2. Amazon Bedrock Access: Ensure your AWS account has access to embedding models like Titan Embeddings
3. Environment Variables: The server uses AWS_REGION and AWS_PROFILE environment variables

Optional Requirements

1. GitHub Token: Set GITHUB_TOKEN environment variable for higher rate limits when searching GitHub repositories

Installation

Kiro	Cursor	VS Code

To add this MCP server to Kiro or Claude, add the following to your MCP config file:

{
  "mcpServers": {
    "awslabs.git-repo-research-mcp-server": {
      "command": "uvx",
      "args": ["awslabs.git-repo-research-mcp-server@latest"],
      "env": {
        "AWS_PROFILE": "your-profile-name",
        "AWS_REGION": "us-west-2",
        "FASTMCP_LOG_LEVEL": "ERROR",
        "GITHUB_TOKEN": "your-github-token"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Windows Installation

For Windows users, the MCP server configuration format is slightly different:

{
  "mcpServers": {
    "awslabs.git-repo-research-mcp-server": {
      "disabled": false,
      "timeout": 60,
      "type": "stdio",
      "command": "uv",
      "args": [
        "tool",
        "run",
        "--from",
        "awslabs.git-repo-research-mcp-server@latest",
        "awslabs.git-repo-research-mcp-server.exe"
      ],
      "env": {
        "FASTMCP_LOG_LEVEL": "ERROR",
        "AWS_PROFILE": "your-aws-profile",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}

Tools

create_research_repository

Indexes a Git repository (local or remote) using FAISS and Amazon Bedrock embeddings.

create_research_repository(
    repository_path: str,
    output_path: Optional[str] = None,
    embedding_model: str = "amazon.titan-embed-text-v2:0",
    include_patterns: Optional[List[str]] = None,
    exclude_patterns: Optional[List[str]] = None,
    chunk_size: int = 1000,
    chunk_overlap: int = 200
) -> Dict

search_research_repository

Performs semantic search within an indexed repository.

search_research_repository(
    index_path: str,
    query: str,
    limit: int = 10,
    threshold: float = 0.0
) -> Dict

search_repos_on_github

Searches for GitHub repositories based on keywords, scoped to AWS organizations.

search_repos_on_github(
    keywords: List[str],
    num_results: int = 5
) -> Dict

access_file

Accesses file or directory contents within repositories or on the filesystem.

access_file(
    filepath: str
) -> Dict | ImageContent

delete_research_repository

Deletes an indexed repository.

delete_research_repository(
    repository_name_or_path: str,
    index_directory: Optional[str] = None
) -> Dict

Resources

repositories://repository_name/summary

Get a summary of an indexed repository including structure and helpful files.

repositories://awslabs_mcp/summary

repositories://

List all indexed repositories with detailed information.

repositories://

repositories://index_directory

List all indexed repositories from a specific index directory.

repositories:///path/to/custom/index/directory

Considerations

• Repository indexing requires Amazon Bedrock access and sufficient permissions
• Large repositories may take significant time to index
• Binary files (except images) are not supported for content viewing
• GitHub repository search is by default limited to AWS organizations: aws-samples, aws-solutions-library-samples, and awslabs (but can be configured to include other organizations)