Discovery Engine MCP server

What it does

Discovery Engine finds statistically validated patterns in tabular data without requiring initial hypotheses. It identifies feature interactions, subgroup effects, and conditional relationships through machine learning and interpretability techniques, then validates findings on held-out data with FDR correction and checks them against academic literature.

• Discover unexpected interactions between variables that affect a target outcome, such as specific combinations of patient attributes that predict high readmission rates.
• Identify novel patterns in your data that differ from what is already reported in academic literature, to surface genuinely new insights.
• Explore large datasets without prior assumptions about which variables matter or how they relate to your target, getting structured findings with effect sizes and p-values.

Tools this server exposes

Tool	Does
discover	Analyzes tabular data to find statistically validated patterns in feature interactions and subgroup effects, returning structured results with conditions, p-val

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.