MCP server for Dremio: governed data access for AI agents
A Model Context Protocol server on top of Dremio that lets AI agents answer natural-language questions over enterprise datasets, with the same permissions and audit trail as a human analyst.
The problem
Business users wanted to ask questions of the data lake in plain language. Giving a language model direct SQL access was a non-starter: no permissions, no audit trail, and queries that could scan the whole lake.
What we built
We built an MCP server that exposes Dremio as tools: schema discovery, governed query execution with row limits and timeouts, and result summaries. Access runs through the user’s own credentials, so Dremio’s permissions apply unchanged, and every query is logged together with the question that produced it. An agent on top translates questions into queries and explains the result.
What changed
Analysts and business users ask questions in natural language and get answers together with the query that produced them. The data team kept full control: same permissions, same audit trail, no new copy of the data. The same server now serves several agents and tools.
Outcomes
- Natural-language questions over governed enterprise data
- Dremio permissions and audit trail apply unchanged
- One server reused by several agents and tools