test_http_mcp

You can run a small Python-based MCP server that exposes example Tools and Prompts over HTTP or via standard input. It’s useful for testing MCP-capable clients and experimenting with quick tool prompts in a controlled environment.

How to use

Start HTTP mode to expose MCP endpoints over a web path and connect with any MCP client. Run the server and then open the provided URL to interact with the MCP server.

To run in HTTP mode, start the app using the runtime command and access the MCP interface at the mounted path.

uv run run-app
# → http://localhost:8000/mcp

Connect via an MCP client over HTTP

Configure your MCP client to point to the local HTTP endpoint and supply the required authorization header if you are simulating a token-protected setup.

{
  "mcpServers": {
    "test_http_mcp": {
      "type": "http",
      "url": "http://localhost:8000/mcp/",
      "headers": {
        "Authorization": "Bearer $TEST_TOKEN"
      }
    }
  }
}

Connect via stdio with a local MCP server

You can run a local MCP server process and connect through standard input/output. Use a client that supports stdio configuration and point it to the server startup command.

{
  "mcpServers": {
    "test_studio": {
      "type": "stdio",
      "command": "uv",
      "args": ["run", "run-stdio"],
      "env": { "AUTHORIZATION_TOKEN": "Bearer TEST_TOKEN" }
    }
  }
}

Available tools

get_weather

Fetches weather information for a given location and optional unit. Example: location input returns weather data as a string.

get_time

Returns the current time as a localized string.

tool_that_access_request

Reads the incoming request headers to demonstrate access to the Authorization header and returns a confirming message.

get_called_tools

Reports which tools have been invoked in the current session.