Skip to content

README.md

Latest commit

 

History

History
37 lines (25 loc) · 1.38 KB

README.md

File metadata and controls

37 lines (25 loc) · 1.38 KB

FastMCP Demo Server + Wasmer

This example shows how to run a Model Context Protocol (MCP) server built with FastMCP on Wasmer Edge.

Demo

https://<your-subdomain>.wasmer.app/ (attach /sse when testing from an MCP-compatible client)

How it Works

main.py defines the MCP application:

  • mcp = FastMCP("Demo") creates a server instance at import time so Wasmer can load main:mcp.
  • Three capabilities are registered:
    • add(a, b) – a simple tool that sums two integers.
    • greeting://{name} – a dynamic resource that returns a personalised greeting.
    • greet_user(name, style) – a prompt template that chooses wording based on the requested style.
  • When run as a script, mcp.run(transport="streamable-http") starts the server using the Streamable HTTP transport expected by Wasmer Edge.

Use these patterns to add your own tools, resources, or prompts before deploying.

Running Locally

python -m venv .venv
source .venv/bin/activate
pip install mcp
python main.py

The server listens for MCP clients on the default port and transport (Streamable HTTP). Point your MCP client (e.g., ChatGPT or Claude Desktop) to the local endpoint.

Deploying to Wasmer (Overview)

  1. Ensure main.py exports the mcp instance (entrypoint main:mcp).
  2. Deploy to Wasmer Edge.
  3. Connect your MCP-enabled client to https://<your-subdomain>.wasmer.app/sse.