Summary

Code generated against the schema you actually have.

The recurring failure in database-adjacent code generation is an invented column name, and no amount of prompt wording fixes it. Serving the real schema corrects the input instead of the output.

What it is

A hosted MCP endpoint at api.schemaflow.dev that serves a cached copy of your PostgreSQL or Supabase schema to AI-IDEs over SSE.

What you get

The full schema returned: tables, columns, relationships, functions, triggers, enums and indexes
The same call narrowed to one kind of object when the whole schema is more than you need
A database analysis covering performance, security and structure
A schema alignment check against best practices, returning specific recommendations rather than a score
One-line SSE configuration for Cursor, Windsurf and VS Code with Cline

Requirements

A SchemaFlow account: connect the database in the dashboard, load the schema, generate an MCP token. The token goes in the SSE URL — https://api.schemaflow.dev/mcp/?token=your-token-here — and there is nothing to install.

Setup effort

One command plus a key — npx -y mcp-remote https://api.schemaflow.dev/mcp/?token=your-token-here, then supply credentials