Put a governed semantic engine between AI and business data.
Foggy Data MCP turns AI analysis requests into model-aware, permission-aware, executable query workflows using TM/QM semantic models, JSON Query DSL, MCP tools, and query evidence.
Semantic Query Consolegovernedcustomer · product · orderDate · salesAmount
The issue is not whether an LLM can write SQL. It is whether SQL is the right AI contract.
Raw SQL prompts expose schema details, blur permission boundaries, and force business metrics into fragile prompt text. Foggy moves those responsibilities into semantic models and the query engine, so AI works through governed business fields instead of ad hoc database internals.
A governed analysis workflow from AI request to query evidence.
Semantic Modeling
Define tables, dimensions, measures, exposed fields, and permission boundaries with TM/QM.
02Model Discovery
AI discovers accessible models and fields instead of reading the entire database schema.
03DSL Query
Intent is expressed as JSON DSL, making validation, logging, reuse, and replay practical.
04Execution Evidence
The engine generates dialect SQL and records DSL, SQL, and result summaries for review.
The v1.0 whitepaper is frozen as the current public contract.
The current public release documents implemented and verified semantic modeling, DSL query, compose analysis, governance, and evidence capabilities. Future capabilities stay in later versions and roadmap documents.