OpenAPI Lambda MCP Server

# MCP Tests Shared test utilities for testing MCP servers across all API packages. ## Overview This package provides reusable testing utilities built on the official MCP Python SDK with SSE transport support. It eliminates the need for manual JSON-RPC request handling and SSE parsing. ## Features - **MCP Client Session**: Easy-to-use async context manager for MCP connections - **Environment Handling**: Utilities for loading and validating environment variables - **Tool Discovery**: Helper functions for finding tools by name ## Installation Add to your API package's dev dependencies: ```toml [project.optional-dependencies] dev = [ "mcp-tests", ] ``` ## Usage ### Basic Example ```python import pytest from pathlib import Path from mcp_tests import ( create_mcp_session, load_environment_from_dotenv, validate_required_env_var, find_tool_by_name, ) # Load environment variables load_environment_from_dotenv(Path(__file__).parent / '.env') @pytest.fixture async def mcp_session(): """Create an authenticated MCP session.""" url = validate_required_env_var("MCP_SERVER_URL") async with create_mcp_session(url) as session: yield session @pytest.mark.asyncio async def test_list_tools(mcp_session): """Test listing available tools.""" # When: Requesting the list of tools tools_response = await mcp_session.list_tools() # Then: The response contains tools assert len(tools_response.tools) > 0 print(f"Found {len(tools_response.tools)} tools") @pytest.mark.asyncio async def test_call_tool(mcp_session): """Test calling a tool.""" # Given: A list of available tools tools_response = await mcp_session.list_tools() user_tool = find_tool_by_name(tools_response.tools, "get_current_user") assert user_tool is not None # When: Calling the tool result = await mcp_session.call_tool(user_tool.name, {}) # Then: The tool returns valid data assert len(result.content) > 0 ``` ## API Reference ### `create_mcp_session(url, headers=None)` Create an MCP client session using SSE transport. **Parameters:** - `url` (str): The MCP server SSE endpoint URL - `headers` (dict, optional): HTTP headers to include in requests **Returns:** - `ClientSession`: An initialized MCP client session **Example:** ```python async with create_mcp_session("https://api.example.com/mcp") as session: tools = await session.list_tools() ``` ### `validate_required_env_var(var_name)` Validate that a required environment variable is set. **Parameters:** - `var_name` (str): Name of the environment variable **Returns:** - `str`: The value of the environment variable **Raises:** - `pytest.fail`: If the variable is not set ### `find_tool_by_name(tools, tool_name)` Find a tool by exact name match. **Parameters:** - `tools` (list[Tool]): List of Tool objects from `list_tools()` response - `tool_name` (str): Exact name of the tool to find **Returns:** - `Tool | None`: The Tool object if found, None otherwise ### `load_environment_from_dotenv(file_path)` Load environment variables from a .env file. **Parameters:** - `file_path` (Path, optional): Path to .env file ## Benefits ✅ **Official SDK**: Uses the official MCP Python SDK ✅ **Type Safety**: Full type hints and IDE autocomplete ✅ **Async Native**: Built on modern async/await patterns ✅ **Reusable**: Share code across all API test suites ✅ **Maintainable**: Changes to MCP protocol handled by SDK updates ## License Apache-2.0