datadog-cost-analyze-agent-mcp

# Datadog Cost Analyze Agent MCP server for end-to-end Datadog billable usage and cost analysis. The server exposes one MCP tool: - `datadog_generate_full_cost_report` This tool runs the full workflow in one call: 1. Fetch billable summary from Datadog 2. Calculate org-level on-demand and committed allocation 3. Generate CSV reports 4. Calculate cost report 5. Return report content as text in the MCP response ## Requirements - Python 3.10+ - Datadog credentials with `usage_read` permission: - `DD_API_KEY` - `DD_APP_KEY` - `DD_SITE` (optional, defaults to `datadoghq.eu`) ## Install ```bash python3 -m venv .venv .venv/bin/python -m pip install -r requirements.txt ``` ## Publish to PyPI Package metadata is configured in `pyproject.toml`. Current package/command names: - Package: `datadog-cost-analyze-agent-mcp` - CLI entrypoint: `datadog-cost-analyze-mcp` ### 1) Build distribution files ```bash .venv/bin/python -m pip install --upgrade build twine .venv/bin/python -m build ``` This creates: - `dist/*.tar.gz` (source distribution) - `dist/*.whl` (wheel) ### 2) Validate package metadata ```bash .venv/bin/python -m twine check dist/* ``` ### 3) Upload to TestPyPI (recommended first) ```bash .venv/bin/python -m twine upload --repository testpypi dist/* ``` ### 4) Upload to PyPI ```bash .venv/bin/python -m twine upload dist/* ``` ### 5) Verify install from PyPI ```bash python3 -m pip install <your-package-name> ``` Use API tokens for authentication: - `TWINE_USERNAME=__token__` - `TWINE_PASSWORD=<pypi-token>` ## Run with uvx After publishing to PyPI: ```bash uvx --from datadog-cost-analyze-agent-mcp datadog-cost-analyze-mcp ``` For local testing before publish: ```bash uvx --from . datadog-cost-analyze-mcp ``` ## Run MCP Server ```bash .venv/bin/python mcp_server.py ``` ## MCP Tool ### `datadog_generate_full_cost_report` #### Arguments - `month` (string, required): Month in `YYYY-MM` format. #### Returns - `ok` (boolean): Success st…