OpenAPI Lambda MCP Server

# Harvest API Testing Guide for running tests against the Harvest MCP server. ## Prerequisites 1. **Install dependencies:** ```bash cd packages/mcp_tests uv pip install -e . cd ../apis/harvest uv pip install -e ".[dev]" ``` 2. **Configure credentials:** Create a `.env` file in `packages/apis/harvest/`: ```bash MCP_SERVER_URL=http://localhost:8080 HARVEST_ACCESS_TOKEN=your_harvest_token HARVEST_ACCOUNT_ID=your_account_id ``` ## Running Tests ### Using Just (Recommended) The `justfile` provides convenient recipes for running tests: #### Local Testing (using .env file) ```bash cd packages/apis/harvest just smoke-test ``` This runs tests against the URL configured in your `.env` file. #### Testing Deployed Stack ```bash just smoke-test HarvestAPIStack ``` This automatically: 1. Fetches the API URL from CloudFormation stack outputs 2. Sets `MCP_SERVER_URL` environment variable 3. Runs tests against the deployed API ### Using pytest Directly You can also run tests directly with pytest: ```bash # From repository root cd packages/apis/harvest source ../../../.venv/bin/activate python -m pytest test_harvest_mcp.py -v -s ``` **Options:** - `-v`: Verbose output - `-s`: Show print statements - `-k test_list_tools`: Run specific test ## Test Structure ### Test Files - `test_harvest_mcp.py` - Main integration tests for Harvest API - `test_list_tools` - Verifies tools are available - `test_get_current_user` - Tests user retrieval endpoint ### Test Architecture Tests use the `mcp_tests` package which provides: - **MCP SDK Integration**: Official MCP Python SDK with SSE transport - **Given-When-Then**: Clear test structure - **Async/Await**: Modern async patterns with `pytest-asyncio` - **Type Safety**: Full type hints using MCP types ### Example Test ```python @pytest.mark.asyncio async def test_list_tools(authenticated_url): # Given: An authenticated MCP server URL async with create_mcp_session(authenticated_url) as session: # When: Requesting the list of available tools tools_response = await session.list_tools() # Then: The response contains a valid list of tools assert len(tools_response.tools) > 0 ``` ## CI/CD Integration The `smoke-test` recipe is designed for CI/CD pipelines: ```yaml # Example GitHub Actions - name: Run Harvest Tests env: HARVEST_ACCESS_TOKEN: ${{ secrets.HARVEST_ACCESS_TOKEN }} HARVEST_ACCOUNT_ID: ${{ secrets.HARVEST_ACCOUNT_ID }} run: | just smoke-test HarvestAPIStack ``` ## Troubleshooting ### Connection Refused ``` httpcore.ConnectError: [Errno 61] Connection refused ``` **Solution**: Ensure the MCP server is running at the URL in `MCP_SERVER_URL`. ### Missing Credentials ``` pytest.fail: HARVEST_ACCESS_TOKEN environment variable not set ``` **Solution**: Configure credentials in `.env` file or set environment variables. ### Import Errors ``` ModuleNotFoundError: No module named 'mcp_tests' ``` **Solution**: Install the `mcp_tests` package: ```bash cd packages/mcp_tests && uv pip install -e . ``` ## Writing New Tests When adding new tests for Harvest API: 1. Use the `mcp_tests` package helpers 2. Follow Given-When-Then pattern 3. Mark tests as `@pytest.mark.asyncio` 4. Use descriptive assertion messages Example: ```python from mcp_tests import create_mcp_session, find_tool_by_name @pytest.mark.asyncio async def test_my_feature(authenticated_url): # Given: An authenticated session async with create_mcp_session(authenticated_url) as session: # When: Calling my tool tools = await session.list_tools() my_tool = find_tool_by_name(tools.tools, "my_tool_name") result = await session.call_tool(my_tool.name, {"arg": "value"}) # Then: The result is valid assert len(result.content) > 0 ``` ## Related Documentation - [MCP Tests Package README](../../mcp_tests/README.md) - [MCP Python SDK](https://github.com/modelcontextprotocol/python-sdk) - [Main Project README](../../../README.md)