Preset MCP: AI That Doesn't Just Answer — It Builds
Today, Preset is launching Preset MCP — now available for Enterprise customers, and with it, a fundamentally different relationship between AI and analytics.
Most AI features in BI tools are chatbots that answer questions about your data. Preset MCP is something different: AI that operates your analytics platform. It creates charts, builds dashboards, executes SQL, explores datasets — all through the same permissions, the same row-level security, and the same RBAC controls your team already relies on.
Here's what we've built.
What Is MCP, and Why Does It Matter?
The Model Context Protocol is an open standard — think of it as a universal interface between AI and external tools. The same way USB-C lets any device talk to any charger, MCP lets any compliant AI client talk to any compliant tool server.
For analytics, this means AI agents can now discover what Apache Superset™ can do, understand your data, and take action — not just generate text responses, but actually build the thing you asked for.
Connect Claude, use the Preset Chatbot, or bring your own agent. The tools are the same. The security is the same. The results are real.
Preset's Role: Building the Foundation in the Open
Preset didn't just add MCP support — we built the open-source foundation it runs on.
The Apache Superset MCP service (proposed in SIP-187 and led by Preset engineers) establishes something most MCP integrations skip: a deep, scalable architecture that makes Superset not just another API-driven tool, but a genuine platform for AI-powered analytics.
The key design choices that make this different:
- Library-first, not API-proxied — the MCP service imports Superset's internal DAOs and models directly, giving AI agents access to the full depth of Superset's capabilities, not just its REST surface
- 20 discrete tools across charts, dashboards, datasets, SQL Lab, and system management — granular enough for agents to reason about each operation separately
- Zero privilege escalation — the same RBAC, row-level security, and column restrictions that govern the web UI apply identically to every AI action
- Preview-first design — agents explore and iterate without polluting the database; charts and dashboards are only persisted when explicitly requested
The result: Superset isn't just the largest open-source analytics platform — it's now one of the most capable AI analytics platforms, period.
Preset Chatbot: Agentic Analytics, Built In
For external power users — engineers, data teams, AI builders — the MCP protocol lets Claude Desktop, Claude Code, Cursor, and custom agents connect directly to your Preset workspace.
For everyone else, there's Preset Chatbot: a built-in conversational AI experience that lives directly inside Superset, no configuration required. Preset Chatbot is currently in beta — reach out to our team to request access or get a demo.
Open it from the floating button in the bottom-right corner of your workspace. Ask it anything:
- "Show me revenue by region for the last 90 days"
- "Build a dashboard for our Q1 executive review"
- "Which datasets have columns related to customer churn?"
- "Run a query comparing this week's signups to last week's"
Preset Chatbot doesn't just answer — it acts. A LangGraph agent orchestrates the response, calling the right MCP tools in sequence, streaming results back in real time, and picking up where you left off in the next session.
It supports multiple LLM providers. It respects your workspace permissions. And it runs on the same 20 MCP tools available to external AI clients — so the capabilities you get in Claude Desktop are the same ones your analysts get in the UI.
What This Means for Superset's Future
Preset has been the primary commercial steward of Apache Superset since Maxime Beauchemin — Superset's creator — founded Preset to build on it. We've contributed millions of lines of code, led major architectural initiatives, and maintained the project through its growth to over 70,000 GitHub stars.
The Superset MCP service is the next chapter of that story. By building the MCP layer in open source — with an extensible architecture, a rich tool catalog, and a clean security model — we're setting Superset up to be the default analytics platform for the AI era.
Any MCP client that works with open-source Superset works with Preset. Bug fixes and new tools contributed by the community benefit Preset customers. And Preset layers additional enterprise capabilities on top — multi-tenant isolation, OAuth 2.0, a built-in LangGraph agent, and production Kubernetes deployment — so the open-source foundation is just the beginning. See what Preset adds →
The ecosystem grows together.
Availability
Preset MCP is now available for Enterprise customers.
Enterprise customers get:
- Full access to all 20 MCP tools via external clients (Claude Desktop, Claude Code, and more)
- Preset Chatbot (built-in conversational AI)
- OAuth 2.0 + PKCE authentication for secure, browser-based AI client connections
- Multi-tenant workspace isolation — each workspace's data stays completely separate
Already a Preset Enterprise customer? Reach out to your account representative to get MCP enabled on your workspace.
Interested in Preset Enterprise? Talk to our team — we'd love to show you what this looks like with your data.
Join Us: Live MCP Demo Webinar
See Preset MCP and Preset Chatbot in action — live, with real data, real agents, and real workflows.
Go Deeper
- Apache Superset MCP Service: A Technical Deep Dive — the complete engineering reference: all 20 tools, the security model, middleware pipeline, and deployment guide
- Preset MCP: From Open Source to Enterprise — how Preset extends the OSS service with multi-tenant isolation, OAuth 2.0, and production Kubernetes deployment
- github.com/apache/superset — source code, issues, and contributions welcome
