Matillion MCP

Matillion's MCP server exposes the Data Productivity Cloud to AI clients through a large tool set split by risk. The read side is always available: list-projects and list-published-pipelines describe what exists, list-pipeline-executions, get-execution-details and get-execution-steps drill from a run down to the failing step, get-consumption and get-consumption-etl-users report credit usage by workload and by user, list-streaming-pipelines and get-streaming-pipeline-status cover streaming state, list-agent-tasks and get-agent-task-messages track delegated agent work, and get-lineage-events and get-audit-events supply lineage and audit history. Write tools are gated behind disabling read only mode, which is the default. Those include execute-pipeline and patch-pipeline-execution to trigger or stop a run, schedule tools to create, update and delete schedules, streaming tools to create a pipeline and send it commands, project and environment creation and deletion, artifact creation and promotion between environments, and task tools that hand work to Matillion's own agents and submit decisions back to them. The server runs through npx on a recent Node.js runtime and authenticates with an API client ID and secret or an existing bearer token, with the deployment region selected in configuration. It fits data engineering teams who want an assistant that can explain why last night's load failed, check which project is consuming credits, and rerun the pipeline once the cause is fixed, plus platform teams promoting artifacts between environments.