AI & MCP

BuildPad DaaS includes a built-in Model Context Protocol (MCP) server that lets AI assistants — Claude, Cursor, and others — query and mutate your data safely through a standardized tool interface.

What is MCP?

The Model Context Protocol is an open standard for connecting AI models to external data sources and tools. MCP provides a unified interface that any compatible client can use without custom integration code.

How It Works


AI Client (Claude / Cursor)
        │
        │  JSON-RPC over HTTP
        ▼
BuildPad DaaS MCP Server (POST /api/mcp)
        │
        │  Internal service calls (auth + RBAC enforced)
        ▼
   PostgreSQL (Supabase)

Every tool call is authenticated — via a static Bearer token or a cookie session — and respects the associated user’s role and permissions. AI agents can only do what the user is authorized to do.

Configuration

Environment variables

Variable	Default	Description
`MCP_ENABLED`	`true`	Enable or disable the MCP server
`MCP_ALLOW_DELETES`	`false`	Allow delete operations via MCP
`MCP_PROMPTS_COLLECTION`	—	Collection storing prompt templates
`MCP_SYSTEM_PROMPT`	—	Custom system prompt override
`MCP_SYSTEM_PROMPT_ENABLED`	`true`	Enable the `system-prompt` tool

Database settings

These settings can also be managed via the DaaS Studio at Settings → AI Settings:

Setting	Type	Description
`mcp_enabled`	boolean	Enable/disable MCP
`mcp_allow_deletes`	boolean	Allow deletes via MCP
`mcp_prompts_collection`	text	Collection for AI prompt templates
`mcp_system_prompt`	text	System prompt injected into every session
`mcp_system_prompt_enabled`	boolean	Enable `system-prompt` tool

Settings API

Method	Path	Description
`GET`	`/api/settings/mcp`	Get MCP settings (admin)
`PUT`	`/api/settings/mcp`	Update MCP settings (admin)
`GET`	`/api/settings/mcp/collections`	List available collections

Keep MCP_ALLOW_DELETES=false in production unless the AI agent specifically needs delete access. Use a dedicated low-privilege MCP user rather than an admin account.

System Prompt

Configure a system prompt to give the AI agent context about your data model, conventions, and what operations it should or should not perform:


{ "mcp_system_prompt": "You are managing a content platform. Do not delete any published articles without explicit confirmation. Always set status='draft' for newly created items." }

The system prompt is returned by the system-prompt tool and injected automatically by MCP-compatible clients that support it.