MCP Server Search and Discovery Tools

Overview

This repository provides a focused set of tools for searching and discovering MCP (Model Context Protocol) servers. It collects or indexes server listings, exposes a simple HTTP API for querying server details, and offers a “best match” endpoint to find the most suitable server given a set of capabilities, tags, or other filters. The project is useful for service registries, client bootstrapping, monitoring dashboards, and developer tooling that needs to locate MCP-capable endpoints programmatically.

Designed for developers who want an easy-to-run directory of MCP servers, the project supports full-text and attribute-based search, configurable ranking, and both ad-hoc and automated discovery workflows. It can be deployed locally for testing or in production behind a standard reverse proxy or load balancer.

Features

• Index and query MCP server listings using HTTP REST endpoints
• Full-text search across server names, descriptions, and tags
• Filter by capabilities, region, tags, or custom metadata
• “Best match” ranking to select the most appropriate server for a requested capability set
• Lightweight Node.js/Express implementation with optional Docker image
• Simple JSON API responses for easy client integration
• Configurable via environment variables (port, database, indexing options)

Installation / Configuration

Clone the repository and use npm (or yarn/pnpm) to install dependencies.

git clone https://github.com/atonomus/mcp-servers-search.git
cd mcp-servers-search
npm install

Development and production commands:

# Run in development (auto-reload)
npm run dev

# Build and start production
npm run build
npm start

Docker:

# Build
docker build -t mcp-servers-search:latest .

# Run (example)
docker run -e PORT=8080 -p 8080:8080 mcp-servers-search:latest

Environment variables (common):

Example .env:

PORT=3000
DATABASE_URL=sqlite://./data/servers.db
INDEX_PATH=./data/index
DATA_REFRESH_INTERVAL=3600

Available Resources

The server exposes a small set of HTTP endpoints suitable for direct consumption by clients, dashboards, or other services.

• GET /api/servers
  • Returns a list of all servers (supports pagination).
• GET /api/servers/:id
  • Retrieve full metadata for a specific server.
• GET /api/search
  • Search by query string, tags, capabilities, region, and other filters.
• GET /api/match
  • Find the best match for a requested capability set or tags.

Sample endpoints table:

Example JSON response (server item):

{
  "id": "srv_123",
  "name": "MCP West Coast",
  "url": "https://west.example.com/mcp",
  "region": "us-west",
  "tags": ["image", "vision", "low-latency"],
  "capabilities": ["image-classification", "image-caption"],
  "score": 0.92,
  "meta": { "uptime": "99.9%" }
}

Use Cases

1. Client bootstrapping

   • A client wants to find a nearby MCP server that supports “image-caption”:
     • Request: GET /api/match?capabilities=image-caption&region=eu
     • Response: Best matching server URL and score for immediate connection.

2. Service dashboard and monitoring

   • A dashboard polls /api/servers to build an overview of available MCP endpoints and displays tags, latency averages, and uptime meta.

3. Automated routing / gateway

   • A gateway queries /api/match with an incoming request’s required capabilities and routes traffic to the selected MCP server, improving client experience by matching capabilities and geography.

4. Developer exploration

   • A developer searches for experimental servers capable of “multimodal” features:
     • Request: GET /api/search?q=multimodal&tags=experimental&limit=10

Concrete example (curl):

# Search for servers with image capabilities and the tag "low-latency"
curl "http://localhost:3000/api/search?q=image&tags=low-latency&limit=5" | jq

# Find the best match for a capability set
curl "http://localhost:3000/api/match?capabilities=image-caption,image-classification&region=us-west"

Extending and Integrating

• Indexing: The search index can be extended to include custom scoring (e.g., ping latency, recent success rates).
• Data sources: Ingest lists from static JSON, admin UIs, or remote registries. The codebase includes hooks for scheduled refresh tasks.
• Security: Deploy behind API gateways or add token-based auth for private registries. Rate limiting and caching layers are recommended for high-traffic environments.

Repository and contribution details are available on GitHub: https://github.com/atonomus/mcp-servers-search.