Sigma MCP

Sigma's MCP server exposes the analytics platform's modeled data and query engine to AI assistants, so an agent can find data without a human first locating the right table. It can list the connections, data sources and workbooks a user has access to, describe the columns and definitions behind a dataset, and run queries that execute against the underlying cloud data warehouse through Sigma rather than around it. Routing queries through Sigma matters because the platform's permissions, row level security and published definitions stay in force. An agent asking about revenue gets the governed definition the analytics team maintains, not a hand-rolled SQL guess, and it can only read what the connecting user is entitled to see. That makes conversational analysis workable for business questions where a raw warehouse connection would be too risky or too easy to get wrong. Typical users are analytics engineers and data teams who publish models in Sigma, plus the analysts and operators who consume them. Common tasks include exploring an unfamiliar dataset, pulling a number for a report, checking how a metric is defined, and finding the workbook where a chart lives. The server is a hosted remote MCP endpoint tied to a Sigma organization, so it connects to Claude and other MCP clients using the user's own Sigma credentials.