MCP Servers on Railway

A monorepo hosting multiple MCP (Model Context Protocol) servers on Railway, each exposed over Streamable HTTP via Supergateway.

Services

Service	Source	Description	Endpoint
brave-search	@modelcontextprotocol/server-brave-search	Web search via Brave Search API	https://<service>.railway.app/mcp
crawl4ai	crawl4ai	Open-source LLM-optimized web crawler & Playwright scraper	https://<service>.railway.app/mcp
google-maps	@modelcontextprotocol/server-google-maps	Location search, directions, and place details via Google Maps API	https://<service>.railway.app/mcp
grubhub	aserper/grubhub-mcp	Search restaurants, browse menus, manage cart, place orders	https://<service>.railway.app/mcp
puppeteer	@modelcontextprotocol/server-puppeteer	Browser automation using headless Chromium	https://<service>.railway.app/mcp

Architecture

Each service is a Docker container that:

1. Installs the underlying MCP server (npm package or pip package)
2. Runs Supergateway to bridge stdio → Streamable HTTP
3. Exposes a /mcp endpoint on Railway's assigned port

┌─────────────────────────────────────────────┐
│  Railway Service                            │
│                                             │
│  Supergateway (HTTP :$PORT)                 │
│       │                                     │
│       └── stdio ──▶ MCP Server              │
│                     (brave-search/crawl4ai/ │
│                      google-maps/grubhub/   │
│                      puppeteer)             │
│                                             │
│  Endpoint: https://svc.railway.app/mcp      │
└─────────────────────────────────────────────┘

Setup

1. Clone & configure

git clone https://github.com/<your-username>/mcps.git
cd mcps
cp .env.example .env
# Edit .env with your API keys

2. Environment Variables

Service	Variable	Required	Description
brave-search	BRAVE_API_KEY	✅	Get from brave.com/developers
crawl4ai	(none)	—	Open-source, runs Playwright locally out-of-the-box
google-maps	GOOGLE_MAPS_API_KEY	✅	Get from console.cloud.google.com
grubhub	GRUBHUB_EMAIL	⚠️	Required for account access, cart & ordering
grubhub	GRUBHUB_PASSWORD	⚠️	Required for account access, cart & ordering
puppeteer	(none)	—	Runs headless Chromium out-of-the-box

3. Deploy to Railway

1. Create a new project on Railway
2. Connect this GitHub repo
3. Add services pointing to their subdirectories:
   • brave-search/ → service name: brave-search
   • crawl4ai/ → service name: crawl4ai
   • google-maps/ → service name: google-maps
   • grubhub/ → service name: grubhub
   • puppeteer/ → service name: puppeteer
4. Set environment variables on each service
5. Railway auto-detects the Dockerfile and deploys

4. Connect to your MCP client

Use the Railway-assigned URLs in your MCP client config:

{
  "mcpServers": {
    "brave-search": {
      "url": "https://brave-search-production-e850.up.railway.app/mcp"
    },
    "crawl4ai": {
      "url": "https://crawl4ai-production-64f5.up.railway.app/mcp"
    },
    "google-maps": {
      "url": "https://google-maps-production-4cff.up.railway.app/mcp"
    },
    "grubhub": {
      "url": "https://grubhub-production.up.railway.app/mcp"
    },
    "puppeteer": {
      "url": "https://puppeteer-production-88ab.up.railway.app/mcp"
    }
  }
}

Adding More MCP Servers

1. Create a new directory (e.g., my-new-mcp/)
2. Add a Dockerfile following the same pattern (install MCP + supergateway, wrap with CMD)
3. Add a new service in Railway pointing to the directory
4. Deploy

License

MIT