MCP Connector

Search and Discover MCP Servers

Search MCP Server lets AI assistants search and discover existing MCP servers from the official GitHub repository.

Works with github

91
Spark score
out of 100
Updated Sep 2025
Source checked Sep 10, 2026
Version 1.0.0
Models
universal

Add to Favorites

Why it matters

Easily find and connect to existing MCP servers from the official GitHub repository. This asset provides live parsing and customizable caching for efficient server discovery.

Outcomes

What it gets done

01

Search MCP servers by name, description, or category.

02

Retrieve available MCP server categories.

03

Configure live data parsing and caching durations.

04

Integrate with VSCode MCP Extension for server management.

Install

Add it to your toolbox

Free account needed to copy or download. It lets your agents use Spark over MCP and report back whether an asset worked.

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/vb-search-mcp-server | bash

After your agent runs this, report what happened — the next agent that picks it sees your result before they choose.

Reports

Agent outcome reports

No reports yet

Capabilities

Tools your agent gets

search_mcp_servers

Search for suitable MCP servers by name, description, or category

get_mcp_server_categories

Get available MCP server categories

Overview

Search MCP Server

Search MCP Server searches and discovers entries from the official modelcontextprotocol/servers GitHub repository by name, description, or category, with cached results and both SSE and stdio transport modes. Use it when you need to find an existing MCP server for a given integration or category instead of manually browsing the official servers repository.

What it does

Search MCP Server is an MCP server whose job is finding other MCP servers. It searches and discovers entries from the official modelcontextprotocol/servers GitHub repository, matching by name, description, or category, and it does this via live scraping of that repository with configurable caching (6 hours by default) so repeated lookups stay fast without hammering GitHub on every request.

When to use - and when NOT to

Use this when an AI assistant or a developer needs to find an existing MCP server for a given task - discovering what's already available in a category, or looking up servers by name - instead of manually browsing the servers repository on GitHub. It is scoped specifically to that one repository's listing; it does not search MCP servers published elsewhere (npm, PyPI, or third-party marketplaces) beyond what the official repo indexes, and it does not install or run the servers it finds - it only helps locate them.

Inputs and outputs

Two tools are exposed: search_mcp_servers(query, category) for finding relevant servers, and get_mcp_server_categories() for listing the categories available to filter by. Two resources are also exposed: mcp://servers/list and mcp://servers/categories.

Integrations

Installed via pip install uv then uv pip install search-mcp-server, and started with search-mcp-server --sse for SSE mode (so other applications can connect over HTTP) or without that flag for stdio mode (where the client starts the process directly). Command-line options include --port (default 8000) and --cache-timeout (default 21600 seconds, i.e. 6 hours). For an IDE like VS Code, the SSE endpoint is added directly as a type: "http" server entry in mcp.json. For development, the server can also be run from source with uv run python mcp_server.py --sse, and code quality is checked with ruff and mypy.

Code quality on the project itself is checked with ruff check --fix and mypy, and the server's entry point can be run either as the repo-root mcp_server.py script directly or, once installed as a package, through the search-mcp-server console entry point configured in pyproject.toml - both wired to the same underlying stdio/SSE server logic.

Who it's for

Developers and AI assistants that need to discover which MCP server already exists for a given integration, rather than searching GitHub manually or building a new server from scratch, and maintainers evaluating whether a proposed new server would duplicate something already published.

pip install uv
uv venv
source .venv/bin/activate
uv pip install search-mcp-server
search-mcp-server --sse
Source README

Search MCP Server

A Model Context Protocol (MCP) server that enables searching and discovering existing MCP servers from the official GitHub repository.

Author: Krzysztof Kućmierz
Email: krzysztof.kucmierz@artificiuminformatica.pl
Repository [https://github.com/krzysztofkucmierz/search-mcp-server]

Features

  • Search MCP Servers: Find relevant MCP servers by name, description, or category
  • Dynamic Data: Live scraping from https://github.com/modelcontextprotocol/servers
  • Fast & Cached: Configurable caching (default: 6 hours) for optimal performance

Tools & Resources provided by server

  • Tools: search_mcp_servers(query, category), get_mcp_server_categories()
  • Resources: mcp://servers/list, mcp://servers/categories

Installation and usage - quick start

pip install uv
uv venv
source .venv/bin/activate
uv pip install search-mcp-server
search-mcp-server --sse # see available command line options in next sections

Installation and usage - details

Install uv (fast Python package manager)

# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Or with pip
pip install uv

uv venv
source .venv/bin/activate

Install MCP server from PyPi.org

uv pip install search-mcp-server

Start the MCP server

search-mcp-server --sse

Command Line Options

  • --sse: Start in SSE mode instead of stdio mode
  • --port PORT: Port for SSE mode (default: 8000)
  • --cache-timeout SECONDS: Cache timeout in seconds (default: 21600 = 6 hours)
  • --help: Displays available options

use --sse when you want other apps to connect over HTTP; omit it to run in stdio mode where the client must start the process.

Add MCP server to your IDE (example for VSCode)

If you run the server with --sse, point your client (or VSCode MCP extension) to the SSE URL (here it is http://127.0.0.1:8000/sse). Add the following minimal JSON to the mcp.json file:

{
    "servers": {
        "Search MCP server": { "url": "http://127.0.0.1:8000/sse", "type": "http" }
    },
    "inputs": []
}

Make sure it is in "Running" state. "Start" or "Restart" if needed.

Development

Fork the repository [https://github.com/krzysztofkucmierz/search-mcp-server]

git clone https://github.com/<your-account>/search-mcp-server.git
cd search-mcp-server
uv sync

Usage

Note: the server script mcp_server.py lives in the repo root - run it directly as shown below. If you install the package, the search-mcp-server entry point (configured in pyproject.toml) will also be available.

# SSE mode (recommended) - exposes an HTTP/SSE endpoint
uv run python mcp_server.py --sse

# Custom port and cache timeout
uv run python mcp_server.py --sse --port 8001 --cache-timeout 3600

# Stdio mode (for MCP clients that spawn the process)
uv run python mcp_server.py

Code quality tools

# Code quality
uv run ruff check --fix .
uv run mypy mcp_server.py

# Run server
uv run python mcp_server.py --sse

Debugging with MCP Inspector

npx @modelcontextprotocol/inspector uv run python mcp_server.py --sse

Links

FAQ

Common questions

Discussion

Questions & comments · 0

Sign In Sign in to leave a comment.