It is a thin, self-hosted bridge — one package, two environment variables, and an API token whose lifetime and permissions you set. What it is documented for is the reporting question: what is deployed, what is being used, and what that means for the next decision.
Waldur's MCP server connects a Waldur instance to Claude Desktop, giving the assistant direct API access for reporting and deeper analysis of resource and usage data.
- Your Waldur instance queried in conversation, for reporting and analysis of resources and usage
- Direct API access that complements the llms.txt documentation files Waldur publishes
- A published-package configuration through uvx, and a directory configuration for running from a checkout
- Debugging through the MCP Inspector, launched with npx @modelcontextprotocol/inspector
Python 3.13 or higher, the uv package manager, and the waldur-mcp-server package from pip or uv. Two environment variables carry the connection: WALDUR_API_URL and WALDUR_TOKEN. The token comes from Username > Credentials > API Token in your instance, with a lifetime you choose.
One command plus a key — pip install waldur-mcp-server, then supply credentials