MCP Setup
Configure the Conduit Admin MCP server for development-time provisioning.
The Conduit MCP server exposes Admin API operations as tools so AI agents can provision schemas, users, and config during development. Never call MCP from application runtime code.
For AI agents
MCP is dev/deploy only. Apps use the Client API at port 3000 with user tokens. See Client vs Admin API.
Prerequisites
- Running Conduit instance with Admin API reachable (default port 3030)
- MCP transport enabled:
transports.mcp: truein admin config - Admin JWT or
cdt_API token
Cursor configuration
Add to .cursor/mcp.json:
{
"mcpServers": {
"conduit": {
"url": "http://localhost:3030/mcp?modules=authentication,database,storage,authorization",
"headers": {
"Authorization": "Bearer YOUR_ADMIN_JWT_OR_cdt_TOKEN"
}
}
}
}Protocol: Streamable HTTP (2025-06-18). Path is fixed at /mcp.
Module loading
On connect, only core and __meta__ load by default. Enable modules via the URL query:
| Need | Add to URL |
|---|---|
| Schemas, custom endpoints | database |
| Users, teams | authentication |
| Files | storage |
| ReBAC | authorization |
| Chat | chat |
| Email/SMS/push | communications (or email,push,sms) |
After changing the URL, reconnect MCP and call list_modules to verify loaded: true.
Authentication
- Admin JWT — from admin login flow
- API token — create via Admin API (
cdt_prefix, shown once)
Every request: Authorization: Bearer <token>.
Tool naming
Admin routes map to tools: GET /database/schemas → get_database_schemas.