Monte Carlo MCP

Monte Carlo's MCP server connects AI assistants to a data observability workspace, so an agent can investigate pipeline problems using the same signals the platform surfaces in its own interface. Through it an assistant can list and read data incidents, pull the monitors and rules that fired, check table health metrics such as freshness, volume and schema changes, and trace upstream and downstream lineage to see which assets a broken table feeds. The workflow it enables is triage. When a dashboard goes stale or a model output looks wrong, an agent can retrieve the relevant incident, correlate it with recent schema or volume anomalies, walk the lineage graph toward the likely source table, and summarize the blast radius for the team. It can also query the catalog for table and field level metadata, which makes it useful for answering ownership and documentation questions that would otherwise mean hunting through the warehouse. It is aimed at data engineers, analytics engineers and data platform owners who already run Monte Carlo over their warehouse and want incident context inside a coding agent or chat client rather than a separate tab. Access uses Monte Carlo API keys, so the agent sees the same accounts and domains that key is authorized for, and the server speaks standard MCP so it drops into common clients without custom glue code.