MCP Servers

Connect your AI assistant directly to your services via Model Context Protocol (MCP).

What are MCP Servers?

MCP (Model Context Protocol) allows AI assistants (like Claude, Cursor, Antigravity) to safely interact with external tools. By configuring these servers, your AI can:

• Read/Write Database: Inspect table schemas and run queries.
• Check Product Analytics: Fetch ordinary lifecycle and onboarding insights from Wire.
• Operate Firebase Services: Work with project configuration, Firestore, Auth, messaging, Crashlytics, and Remote Config.
• Manage Subscriptions: Check RevenueCat customers.
• Read Designs: Fetch component specs from Figma.

Available Servers

• Supabase MCP: Database schema and queries.
• Firebase MCP: Project setup, Firestore, Auth, messaging, Crashlytics, and Remote Config.
• Wire AI MCP: Product events, onboarding insights, experiments, and learning reports.
• RevenueCat MCP: Subscription data.
• Context7 MCP: Enhanced codebase context.
• Figma MCP: Design-to-code integration.

Configure the active coding host

The boilerplate does not ship a .mcp.json. Follow the linked guide for each service because command, transport, and authentication differ. For Claude Code, prefer the documented claude mcp add ... command and verify with claude mcp list. For Cursor or Antigravity, use that client's MCP settings and the exact command or hosted URL from the service guide.

Keep credentials in local environment or host secret settings. Record a successful, secret-free connection receipt in docs/KEYS.md; choosing a server during initialization is only a plan.