mcp-transport-prototype

What is mcp-transport-prototype?

mcp-transport-prototype is a prototype implementation for understanding how to build a custom MCP server accessible over both stdio and SSE transports. It explores the differences between Server-Sent Events (SSE) and standard input/output (stdio) for MCP communication.

How to use mcp-transport-prototype?

Run the server with uv run main.py in one terminal, and the client with uv run mcp_client.py in another to test the SSE transport. No further configuration details are provided.

Key features of mcp-transport-prototype

• Implements MCP server over SSE and stdio transports
• Demonstrates one-way real-time streaming via SSE over HTTP
• Shows traditional process communication via stdio
• Includes a sample client script to test SSE mode

Use cases of mcp-transport-prototype

• Understanding how to build custom MCP servers with different transports
• Testing SSE for web dashboards or browser-based apps needing live updates
• Using stdio for command-line tools or LLM agents launching MCP as subprocess

FAQ from mcp-transport-prototype

What is the difference between SSE and stdio transports?

SSE runs the MCP server as a separate HTTP service, pushing real-time updates one-way from server to client over the network. Stdio launches the server as a subprocess communicating via stdin/stdout, ideal for in-process tools without network overhead.

What runtime dependencies are required?

The prototype uses uv for running Python scripts. No additional dependencies are listed.

How do I test the SSE transport?

Run uv run main.py in one terminal and uv run mcp_client.py in another. The client script is available at the provided GitHub link.

Where does the data live?

The README does not specify data storage; it focuses on transport implementation only.

What are the known limitations?

SSE is one-way (server to client), unlike WebSockets. No other limitations are mentioned.