FastMCP Template

This is a FastMCP template project that works seamlessly in local, Docker, and cloud environments. Get started instantly by deploying to Railway with one click!

🚀 Quick Deploy to Railway (Recommended)

The fastest way to get your FastMCP server running in the cloud:

One-Click Deploy

Deploy on Railway

Connect Remote Client to Railway

Once deployed, use this client to connect from anywhere:

Simple Method (Direct URL):

# my_client_remote.py
import asyncio
from fastmcp import Client

# Replace with your Railway URL
RAILWAY_URL = "https://your-railway-url.up.railway.app/mcp"
client = Client(RAILWAY_URL)

async def call_tool(name: str):
    async with client:
        result = await client.call_tool("greet", {"name": name})
        print(result)

if __name__ == "__main__":
    asyncio.run(call_tool("Ford"))

Run it:

uv run my_client_remote.py

With Environment Variable (More Flexible):

export RAILWAY_URL="https://your-railway-url.up.railway.app/mcp"
uv run my_client_remote.py

---

Project Structure

• my_server.py - FastMCP server with a greet tool (works locally, Docker, and Railway)
• my_client.py - Local/Docker client that connects via HTTP
• my_client_remote.py - Remote client for Railway connections
• Dockerfile - Container configuration for all cloud deployments

Local Development

Perfect for testing and development on your machine.

Terminal 1 - Start the server:

uv run fastmcp run my_server.py:mcp --transport http --port 8080

Terminal 2 - Run the client:

export PORT=8080
export HOST_URL="http://localhost"
uv run my_client.py

Output: Hello, Ford!

Docker Deployment

Deploy locally with Docker or on any container platform.

Build the Docker image:

docker build -t fastmcp-server .

Run the Docker container:

docker run -p 8080:8080 fastmcp-server

Connect the client:

export PORT=8080
export HOST_URL="http://localhost"
uv run my_client.py

Environment Variables

Local & Docker Deployments

• HOST_URL - The server host URL (default: http://localhost)
• PORT - The server port (default: 8080)

Examples:

# Custom port
export PORT=3000
uv run fastmcp run my_server.py:mcp --transport http --port 3000
export PORT=3000
uv run my_client.py

# Custom host
export HOST_URL="http://192.168.1.100"
export PORT=8080
uv run my_client.py

Railway Deployment

• RAILWAY_URL - Full Railway endpoint URL (e.g., https://your-url.up.railway.app/mcp)
• Configure additional variables in Railway dashboard → Variables tab

Architecture

Why This Approach?

This template uses HTTP transport for consistency across all deployment scenarios:

1. Consistency - Same protocol everywhere
2. Simplicity - Just two files for any scenario
3. Scalability - HTTP enables cloud deployment
4. Flexibility - Easy to modify URLs/ports per environment

Deployment Options

┌──────────────────────────────────────────────┐
│  🚀 RECOMMENDED: Railway Cloud                 │
│  • One-click deployment                        │
│  • Automatic HTTPS & CDN                       │
│  • Global access                               │
│  URL: https://your-url.up.railway.app/mcp     │
└──────────────────────────────────────────────┘

┌──────────────────────────────────────────────┐
│  🐳 Docker (Local or Any Cloud)               │
│  • Full control                                │
│  • Works anywhere with Docker                  │
│  URL: http://localhost:8080/mcp               │
└──────────────────────────────────────────────┘

┌──────────────────────────────────────────────┐
│  🏠 Local Development                         │
│  • Perfect for testing                        │
│  • Two terminal setup                         │
│  URL: http://localhost:8080/mcp               │
└──────────────────────────────────────────────┘

Configuration Details

Server (my_server.py)

The server uses the FastMCP framework with a simple greet tool:

@mcp.tool
def greet(name: str) -> str:
    return f"Hello, {name}!"

When run locally with:

uv run fastmcp run my_server.py:mcp --transport http --port 8080

When run in Docker via Dockerfile:

CMD ["sh", "-c", "uv run fastmcp run my_server.py:mcp --transport http --host 0.0.0.0 --port $PORT"]

Client (my_client.py)

Local/Docker client connects with environment variables:

PORT = os.getenv("PORT", "8080")
HOST_URL = os.getenv("HOST_URL", "http://localhost")
client = Client(f"{HOST_URL}:{PORT}/mcp")

Remote Client (my_client_remote.py)

Railway/remote client connects with direct URL or environment variable:

RAILWAY_URL = os.getenv("RAILWAY_URL", "https://your-url.up.railway.app/mcp")
client = Client(RAILWAY_URL)

Troubleshooting

Railway Connection Issues

Server not responding:

1. Check Railway deployment logs in dashboard
2. Verify URL: https://your-url.up.railway.app/mcp
3. Test with curl: curl https://your-railway-url.up.railway.app/mcp

Finding your Railway URL:

1. Go to Railway project dashboard
2. Select the deployment
3. Go to "Settings"
4. Copy the domain URL
5. Append /mcp for the endpoint

Custom Domain on Railway:

• Go to Service Settings → Custom Domain
• Add your domain (e.g., mcp.example.com)

Local Development Issues

Connection failed:

1. Check if server is running: ps aux | grep fastmcp
2. Verify port accessible: curl http://localhost:8080/mcp
3. Ensure PORT and HOST_URL environment variables are set

Port already in use:

# Use different port
export PORT=3000
uv run fastmcp run my_server.py:mcp --transport http --port 3000
export PORT=3000
uv run my_client.py

Docker Issues

1. Check if container is running: docker ps
2. View logs: docker logs <container-id>
3. Verify port mapping: docker run -p 8080:8080 ...
4. Test connectivity: curl http://localhost:8080/mcp

Summary

This FastMCP template provides multiple deployment options:

• ✅ Railway - Fastest cloud deployment (recommended)
• ✅ Docker - Full control, works anywhere
• ✅ Local - Perfect for development and testing
• ✅ Same codebase - No changes needed for different deployments
• ✅ Just change URLs - Environment variables handle all variations