mcp -template

This MCP server is a minimal FastMCP implementation that runs a Python-based MCP server and exposes a Streamable HTTP transport. It serves an MCP endpoint at /mcp and can be interacted with using the included inspector tool to test or debug agent interactions. Start the server with the Python script, then connect a client to the specified MCP transport endpoint to begin exchanging messages.

To explore its capabilities, run the local server and use the inspector (npx @modelcontextprotocol/inspector) to inspect interactions in real time. Open the UI at http://localhost:3000 and point it at the MCP endpoint at http://localhost:8000/mcp using the Streamable HTTP transport. This setup allows you to test tool invocation, message formatting, and response handling in a lightweight, local environment.

Prerequisites:

• Conda (optional but used in the template's setup steps) or a Python 3.13 environment
• Python 3.13
• Git

Step-by-step installation:

1. Clone the repository:

git clone <your-repo-url>
cd mcp-server-template

2. Create and activate the environment (via Conda as shown in the template):

conda create -n mcp-server python=3.13
conda activate mcp-server

3. Install Python dependencies:

pip install -r requirements.txt

4. Run the MCP server locally:

python src/server.py

5. In another terminal, start the inspector client to test interactions:

npx @modelcontextprotocol/inspector

6. Open a browser and navigate to http://localhost:3000, then connect to http://localhost:8000/mcp using the Streamable HTTP transport (note the /mcp path).

Tips and common considerations:

• The server uses a minimal FastMCP setup with a Streamable HTTP transport. Ensure you access the MCP endpoint at /mcp (not the root URL).
• You can extend the server by decorating functions with @mcp.tool to add new tools, as shown in the Customization example in the README.
• If you plan to integrate with Poke, you can follow the Poke setup guidance in the README to connect your MCP server to Poke’s integrations.
• If you rename the project or connections, you may need to clear Poke history to re-establish proper tool invocation order.
• Ensure network accessibility on the ports used (3000 for UI, 8000 for MCP transport) when running in a container or deployment environment.