home / mcp / grep app mcp server
Provides cross-repo code search and retrieval from public GitHub repositories via the grep.app API.
Configuration
View docs{
"mcpServers": {
"ai-tools-all-grep_app_mcp": {
"url": "http://localhost:8603/mcp"
}
}
}You can use this MCP server to search and fetch code from millions of public GitHub repositories with fast, flexible queries. It exposes HTTP and STDIO interfaces so you can integrate it with Claude Code or run it locally for rapid code discovery and learning from open source projects.
You connect to the server from your MCP client either over HTTP or by running a local STDIO instance. Use the HTTP endpoint to perform streaming-enabled MCP requests, or run the STDIO server for direct integration with your tooling. Typical workflows include searching for code patterns, fetching individual files, and batch retrieving multiple files from results.
Prerequisites: verify you have Node.js 18 or newer and npm installed.
1. Clone the project and navigate into the directory.
2. Install dependencies.
3. Build the project.
4. Run the server in HTTP or STDIO mode depending on your needs.
HTTP mode provides a remote MCP endpoint that you can query from your MCP client. The endpoint is ready to use after you start the server.
STDIO mode lets you integrate directly with your tooling by using a local process. The example command starts the STDIO server and exposes an interface for MCP requests.
Logging is enabled with daily rotation and stores logs in the logs/ directory for easy troubleshooting.
The server can be connected in two ways. Use the HTTP endpoint when you want a remote, streaming MCP interface. Use the STDIO interface for tight integration with your local tooling.
Search code across public GitHub repositories with flexible filtering and multiple output formats.
Fetch a single file from a GitHub repository using owner, repo, path, and optional ref.
Fetch multiple files from GitHub repositories in parallel.
Retrieve files from previously cached search results by specifying the original query and result indices.