MCP · AI Autofill · DataTalk

The backend an agent can't screw up.

Let AI agents read, write, and manage your data — through an MCP server with real tools, packaged agent skills, and authorization the agent can't forget, because Postgres enforces it on every query.

Model Context Protocol

MCP Server

The @rebasepro/mcp package exposes your entire Rebase project to AI assistants like Cursor, Claude, and Windsurf via the Model Context Protocol standard.

AI agents can autonomously manage your data, run migrations, generate SDKs, and control the dev server — all through a standardized, secure interface.

Configuration
// .cursor/mcp.json
{
  "mcpServers": {
    "rebase": {
      "command": "npx",
      "args": ["@rebasepro/mcp"],
      "env": {
        "REBASE_BASE_URL": "http://localhost:3001"
      }
    }
  }
}

Available Tools

📝 Data Tools

List, get, create, update, and delete documents in any collection with filtering and pagination.

👤 Admin Tools

List users, create users, assign roles, delete users, and list roles — all programmatically.

🗄️ Schema & DB Tools

schema_generate, db_push, db_pull, db_generate, db_migrate, generate_sdk

🖥️ Dev Server

Start, stop, and read logs from the dev server without leaving your AI assistant.

📁 Resources

Collection definitions and generated Drizzle schema exposed as MCP resources for context-aware AI assistance.

Schema-Aware AI

AI Data Enhancement

The data enhancement plugin analyzes your collection schema and uses AI to auto-fill fields, suggest descriptions, generate SEO metadata, and translate content — all from within the admin UI.

  • ✓ One-click autofill based on existing field values
  • ✓ Schema-aware prompts (knows your field types and constraints)
  • ✓ Inline AI completions in the rich text editor
  • ✓ Works with any OpenAI-compatible API
import { useDataEnhancementPlugin } from
  "@rebasepro/plugin-ai";

const plugin = useDataEnhancementPlugin({
  getConfigForPath: ({ path }) => ({
    suggestions: true,
    autofill: true,
  })
});

// Drop into your Rebase app:
<Rebase
  plugins={[plugin]}
  ...
/>
Natural Language → SQL

DataTalk

Ask questions about your data in plain English. DataTalk converts natural language to SQL, executes it against your Postgres database, and renders the results in charts and tables.

  • ✓ Schema-aware query generation
  • ✓ Auto-generated visualizations
  • ✓ Conversation history and follow-ups
  • ✓ Integrated as a Rebase admin view
You

Show me the top 10 users by order count this month

AI

Here are the top 10 users:

SELECT u.display_name, COUNT(o.id)
FROM users u JOIN orders o ON ...
WHERE o.created_at >= ...
GROUP BY u.id ORDER BY count DESC
LIMIT 10;

Works With Your AI Stack

Rebase integrates with the tools developers are already using.

Cursor

MCP integration

Claude Desktop

MCP integration

Windsurf

MCP integration

OpenAI API

Data enhancement

Agent infrastructure

Everything an agent needs to operate

Vector search, scoped keys, streaming functions, realtime and scheduled tasks — the operational surface agents build on.

Vector Similarity Search

Native pgvector integration with cosine, L2, and inner product distance. Query embeddings via REST API — no external vector DB needed.

GET /api/docs
  ?vector_search=embedding
  &vector=[0.1,0.2,...]
  &vector_distance=cosine

Scoped Service API Keys

Issue keys with granular per-collection, per-operation permissions. Built-in rate limiting, automatic expiry, and Studio management.

SHA-256 hashed Per-key rate limits Soft revocation

Custom Functions & Streaming

Define server-side endpoints with full access to your typed data layer. Stream LLM responses, chain agent actions, or build webhooks — all in TypeScript.

export default async (req, res) => {
  // Full typed access to driver
  const docs = await data.docs.find();
}

MCP Server

Expose your entire project to Cursor, Claude, and Windsurf via the Model Context Protocol. Agents can CRUD data, run migrations, and manage schemas.

Data tools Schema tools Admin tools

Realtime Engine

Full WebSocket engine with live data subscriptions, broadcast channels for messaging, and presence tracking for online status — agents and clients stay in sync instantly.

WebSocket Data subscriptions Broadcast channels Presence tracking Auto-reconnect

Cron Jobs & Background Tasks

Schedule recurring agent tasks with cron expressions. Sync external APIs, process embedding batches, clean stale data, or trigger LLM pipelines on a timer.

Cron expressions Typed handlers Execution logs

Make your data AI-ready

Client SDK Backend & APIs