SSE-based Server and Client for MCP
This demonstrates a working pattern for SSE-based MCP servers and standalone MCP clients that use tools from them. Based on an original discussion here.
Note: Make sure to supply ANTHROPIC_API_KEY in .env or as an environment variable.
uv run weather.py
uv run client.py http://0.0.0.0:8080/sse
Initialized SSE client...
Listing tools...
Connected to server with tools: ['get_alerts', 'get_forecast']
MCP Client Started!
Type your queries or 'quit' to exit.
Query: whats the weather like in Spokane?
I can help you check the weather forecast for Spokane, Washington. I'll use the get_forecast function, but I'll need to use Spokane's latitude and longitude coordinates.
Spokane, WA is located at approximately 47.6587° N, 117.4260° W.
[Calling tool get_forecast with args {'latitude': 47.6587, 'longitude': -117.426}]
Based on the current forecast for Spokane:
Right now it's sunny and cold with a temperature of 37°F and ...
This means the MCP server can now be some running process that agents (clients) connect to, use, and disconnect from whenever and wherever they want. In other words, an SSE-based server and clients can be decoupled processes (potentially even, on decoupled nodes). This is different and better fits "cloud-native" use-cases compared to the STDIO-based pattern where the client itself spawns the server as a subprocess.
weather.py is a SSE-based MCP server that presents some tools based on the National Weather Service APIs. Adapted from the MCP docs' example STDIO server implementation.
By default, server runs on 0.0.0.0:8080, but is configurable with command line arguments like:
uv run weather.py --host <your host> --port <your port>
client.py is a MCP Client that connects to and uses tools from the SSE-based MCP server. Adapted from the MCP docs' example STDIO client implementation.
By default, client connects to SSE endpoint provided in the command line argument like:
uv run weather.py http://0.0.0.0:8080/sse