Let AI query your warehouse, safely
MCP is the common language between AI assistants and the outside world. D23 acts as an MCP server so Claude, GPT, and any AI tool can query your governed Superset semantic layer — with row-level security and a full audit trail on every call.
D23 as an MCP server
A workspace-scoped MCP endpoint exposes your Superset dashboards, datasets, and semantic layer to any MCP client — including Claude.ai, Cursor, and your own apps.
AI as an MCP client
AI assistants connect to D23 over MCP and query your data warehouse in plain language. D23 enforces access controls and translates to SQL — no raw credentials ever reach the AI.
How the MCP layer works
- 1
Connect your warehouse
Add your database or Superset semantic layer to D23 as an MCP-accessible data source.
- 2
Set permissions
Choose exactly which datasets, metrics, and dimensions AI may query — scoped per user or role.
- 3
AI assistants query safely
Claude, GPT, or any MCP client can ask questions in plain language. D23 translates to SQL and enforces access controls.
- 4
Governed end to end
Every AI query runs under row-level security with a full audit trail — no raw warehouse credentials exposed.
What teams use MCP for
Text-to-SQL for your whole team
Slack, Claude.ai, or any MCP client can query your warehouse by asking plain-language questions — D23 handles the SQL.
Semantic layer as an MCP tool
Expose your Superset metrics and dimensions to AI so every answer uses your governed business definitions.
D23 as an MCP server
A workspace-scoped MCP endpoint exposes your dashboards, datasets, and semantic layer to any MCP client.
Security that never moves
Row-level security and audit logs apply to every AI query, the same as any human-initiated dashboard view.
Connect the databases and AI tools you already use
50+ databases, SaaS tools, and AI models — all through one governed MCP layer.
Turn every team into a data team.
Get managed Apache Superset, the dashboards your business needs, and AI on top, without hiring a data team.