AI Integration (MCP)

Let AI assistants understand and interact with your backend directly. Claude, GitHub Copilot, and Cursor can query your data, create records, and manage your application — all through natural language.

57
MCP Tools
2
MCP Options
MCP
Protocol Standard
10x
Dev Speed

Your Backend, AI-Enabled

Model Context Protocol (MCP) is the new standard for connecting AI assistants to external tools and data sources.

Baasix includes a complete MCP server that exposes your entire backend to AI assistants. Choose between Remote MCP (built-in HTTP endpoint at /mcp) for production and cloud deployments, or Local MCP (@baasix/mcp npm package) for Claude Desktop and local development. Instead of switching between your IDE and browser to test APIs, just ask Claude or Copilot to "create a new user named John" or "show me all orders from last week". The AI understands your schema, knows your API structure, and can perform operations on your behalf.

  • 57 built-in tools covering all Baasix operations
  • Remote MCP: Built-in HTTP endpoint for production deployments
  • Local MCP: npm package for Claude Desktop & local dev
  • Schema-aware: AI knows your collections and relationships
  • Works with Claude Desktop, VS Code Copilot, Cursor, and more
  • Natural language queries and data manipulation
  • Permission-aware: AI respects your access control rules

AI-Powered Development

Here's what becomes possible when your AI assistant understands your backend.

✓ With Baasix

  • "Show me users who signed up this week" → instant results
  • "Create a new product with price $29.99" → done
  • "Add a role called moderator with read permissions" → configured
  • "What schema does the orders collection have?" → explained
  • "Update all overdue tasks to status:urgent" → bulk update

✗ Traditional Approach

  • Open browser, navigate to admin, filter users, export
  • Open API client, find endpoint, construct JSON, send
  • Navigate to settings, find permissions, click through UI
  • Find documentation, search for schema definition
  • Write script, test it, run it, verify results

MCP Capabilities

57 Built-in Tools

Query items, create records, manage schemas, handle auth, upload files — every API operation is available to AI.

Remote MCP

Built-in HTTP endpoint at /mcp. Perfect for production, cloud, and remote server deployments.

Local MCP

Standalone npm package for stdio transport. Ideal for Claude Desktop and local development.

Code Generation

Ask AI to generate SDK code for common operations. Get type-safe snippets ready to paste into your app.

Multi-Platform

Works with Claude Desktop, GitHub Copilot in VS Code, Cursor, and any MCP-compatible AI tool.

Extensible

Add custom tools to the MCP server. Expose your business logic to AI assistants.

Transform Your Development Workflow

AI integration accelerates every phase of development.

Rapid Prototyping

Describe your data model in natural language and let AI create schemas, seed data, and test queries for you.

PrototypingSchema DesignTesting

Data Exploration

Ask questions about your data: "What's our most popular product?" or "How many users signed up last month?"

AnalyticsQueriesReporting

Admin Operations

Perform admin tasks without navigating UIs: "Reset password for user@example.com" or "Disable the beta feature flag."

AdminOperationsManagement

Learning & Debugging

Ask AI to explain why a query isn't working, suggest optimizations, or help you understand API responses.

DebuggingLearningSupport

AI Integration FAQ

What is the difference between Remote and Local MCP?

Remote MCP is built into the Baasix server at /mcp endpoint (HTTP transport) — ideal for production and cloud. Local MCP is the @baasix/mcp npm package (stdio transport) — ideal for Claude Desktop and local development.

Is my data sent to AI providers?

Only the data you explicitly query is shared with the AI. The MCP server acts as a bridge — it receives AI requests and returns results. Your data stays on your server.

How do I set up MCP with Claude Code?

Create a .mcp.json file in your project root with the Remote MCP configuration (HTTP type, URL to your /mcp endpoint, and authentication headers).

Does AI have admin access?

The MCP server uses the credentials you configure. You can create a limited role for AI access or use your admin credentials for full control.

Ready to build faster?

Join developers who are shipping production-ready backends in hours, not weeks.