API Reference

MCP Tools Private Preview

RTIE exposes enabled module APIs as Model Context Protocol (MCP) tools at mcp.rtie.ai. Tool availability is tenant-scoped and filtered by workspace type, module enablement, key/session scope, and human-confirmation policy.

Private-preview connection

Use this configuration only if your tenant has MCP preview access.

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "rtie": {
      "url": "https://mcp.rtie.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_RTIE_API_KEY"
      }
    }
  }
}

Agent request shape

const res = await fetch("https://mcp.rtie.ai/mcp", {
  method: "POST",
  headers: {
    "Authorization": "Bearer rtie_live_...",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    jsonrpc: "2.0",
    id: 1,
    method: "tools/call",
    params: {
      name: "list_rooms",
      arguments: { status: "active" }
    }
  }),
});

Tool catalog

The public tool reference is generated from module manifests and the MCP registry. See AI Tools → MCP Private Preview for the current generated list, access levels, and confirmation hints.

Authentication

Use an RTIE API key for tenant/server tools. Runtime agents and user-approved flows should use short-lived RTIE module/session tokens from POST /v1/runtime/connect.

Supabase Auth is a hosted/control-plane adapter, not a runtime requirement. Third-party apps can resolve Auth0, Firebase, Cognito, custom OIDC, or guest actors into RTIE actors before minting session tokens.

Runtime handshake

const handshake = await fetch("https://api.rtie.ai/v1/runtime/connect", {
  method: "POST",
  headers: {
    "X-API-Key": "rtie_live_...",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    roomId: "room_...",
    modules: ["rooms", "presence", "chat"],
    actor: { type: "agent", id: "agent:host", identityProvider: "custom-oidc" }
  }),
}).then((res) => res.json());

Discovery

MCP clients that support auto-discovery will fetch these endpoints once the public server is live:

curl https://mcp.rtie.ai/.well-known/mcp
curl https://mcp.rtie.ai/.well-known/mcp/manifest.json

Module discovery lives at GET https://api.rtie.ai/v1/modules/catalog. Authenticated calls include tenant enablement status; unauthenticated calls return public module metadata.