Data · Internal product
MCP products for a data platform
MCP servers that give AI assistants controlled access to Airflow, Trino, Spark and YARN to operate and query the platform from the chat.
The challenge
Running a data platform with Airflow, Trino and Spark on YARN means jumping between UIs, logs and queries to answer simple questions: what failed, why, and what data is available. That knowledge tends to sit with just a few people.
The solution
We are building a family of MCP servers that expose the platform to AI assistants safely:
- Airflow: DAG and task status, log retrieval and re-runs that require confirmation.
- Trino: catalog exploration and read-only queries in natural language.
- Spark and YARN: application status, resource usage and cluster queues.
Every tool has scoped permissions and every call is logged.
What’s next
The next step is combining these servers into agents that detect an issue, gather context across tools and propose a fix for a person to approve.
Have a process AI could handle?
Tell us what you want to achieve. In a 30-minute call we'll tell you whether it makes sense, how we'd approach it and what to expect.