MCP server

Mavera ships a native Model Context Protocol server, so agent tools can drive the platform directly — no curl scripts, no wrapper code.

{
  "mcpServers": {
    "mavera": {
      "type": "http",
      "url": "https://mavera-platform.vercel.app/api/mcp",
      "headers": { "x-api-key": "YOUR_API_KEY" }
    }
  }
}

That's the whole setup. Your agent gets the curated tool surface (the same registry tools flagged for MCP): audience building, focus groups, deep interviews, surveys, hallway tests, budget planning, insight mining, and the read tools.

What a session looks like

Tell your agent something like:

"Use mavera to test whether 'fresh in 5 minutes' or 'never run out' lands better with busy office managers."

The agent will chain build_audience_from_description → run_study (or a hallway_test first if it's being frugal) and read back the computed aggregates. Every number is code-computed; every call meters credits exactly like the REST API; failed operations refund.

Protocol notes

• Transport: streamable HTTP (stateless), JSON-RPC 2.0 — initialize, tools/list, tools/call
• Auth: your normal API key in the x-api-key header (the same key as the REST API)
• Tool inputs are published as JSON Schema straight from the platform's own validation schemas, so agents see required fields, defaults, enums, and bounds exactly as the engine enforces them
• Long tools (full studies) can take minutes — agents should use sensible timeouts

GET /api/mcp returns a small discovery document if you want to probe the endpoint by hand.