searchCodeLABSCO SUMMARY
Find who else solved this, then read their code.
Search alone gives you filenames; this pairs grep.app's index with file retrieval and caches the results in between, so "show me three real implementations of this pattern" ends with the actual source in front of you.
What it is
An MCP server over the grep.app API for code search across public GitHub repositories, with file retrieval attached so a hit can be opened rather than just cited.
What you get
- Code search across millions of public repositories, with regex, whole-word and case-sensitive matching available
- Filtering by language, by repository and by file path, so the results narrow to the ecosystem you are actually in
- Individual files fetched from GitHub, and batches of files retrieved in parallel
- Result caching, so retrieving files after a search does not re-run the query
- Output as JSON, as a numbered list, or as formatted text
- Two transports: STDIO for direct client integration, or HTTP with streaming
Requirements
Node.js 18+, then clone, npm install and npm run build. A run script starts it in STDIO or HTTP mode, development or production. No account and no key — grep.app indexes public repositories.
Setup effort
Build from source — git clone https://github.com/ai-tools-all/grep_app_mcp.git
USE CASES
01
Finding a real example of an API you are about to use
Search for the call across public repositories, filtered to your language, and read how working projects actually invoke it.
02
Checking whether a pattern is common or a mistake
Look at the spread of matches across repositories before adopting something you found in one blog post.
03
Reading several implementations side by side
Retrieve the matching files in a batch rather than one at a time, so the comparison happens in one pass.
04
Tracking down a cryptic error string
Search the exact message across public code to find the library that emits it and the condition that triggers it.
05
Narrowing a search that returned too much
Add language, repository or path filters, or switch to whole-word and case-sensitive matching, until the result set is readable.
4 tools
Listed in project docs. Your client may show a different list.
github_filegithub_batch_filesbatch_retrieve_filesBEFORE YOU INSTALL
What to know first
Tools you get
4 tools named in the project's documentation.
Requires
Node.js 18+
Where it runs
Both documented setups run the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
git clone https://github.com/ai-tools-all/grep_app_mcp.git
cd grep_app_mcp
npm install
npm run buildMCP Clients
Claude Code — stdio
CLAUDE CODE — STDIO · JSON
{
"mcpServers": {
"grep_app": {
"command": "node",
"args": [
"<YOUR_PATH_TO_GREP_APP_MCP>/dist/server-stdio.js"
],
"env": {}
}
}
}Claude Code — HTTP
CLAUDE CODE — HTTP · JSON
{
"mcpServers": {
"grep_app": {
"type": "http",
"url": "http://localhost:8603/mcp"
}
}
}Start the server first from your clone with ./run.sh http prod; it serves http://localhost:8603/mcp.
