MCP Overview

What is Semaphor MCP?

Semaphor provides a Model Context Protocol (MCP) server that lets AI assistants securely discover and query your analytics data. Connect your SaaS backend, or tools like Claude and Cursor, to ask natural language questions about your data — with all your existing security policies enforced automatically.

The MCP endpoint is available at:

https://semaphor.cloud/api/mcp

Two Integration Patterns

Semaphor MCP supports two ways to connect, depending on your use case.

Embedded MCP

For SaaS applications that want to give their AI agents access to Semaphor data. Your backend generates a project token that scopes the AI agent to a single project — no project selection needed.

• Authenticate with a project token (generated server-side)
• Project scope is pre-determined by the token
• Supports both organization users and tenant users

Best for: Embedded agents within SaaS products, existing applications that want to add Semaphor-powered analytics to their AI features.

Interactive MCP

For organization users who want to connect their personal AI assistant to Semaphor. Sign in via OAuth, discover your projects, then pass projectId on each tool call.

• Authenticate with OAuth (sign in with your Semaphor account)
• Discover projects, then pass projectId on each tool call
• Organization users only

Best for: Claude, Cursor, and other MCP-compatible clients.

What Can You Do?

• Discover — List projects, dashboards, semantic domains, datasets, and schemas
• Query — Run structured analytics and semantically grounded SQL / Python against your data
• Inspect — Get dashboard details, dataset columns, and domain relationships

All operations respect your organization's security policies: connection-level security (CLS), schema-level security (SLS), and row-level security (RLS).

Getting Started