Setting Up MCP Server

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI coding assistants interact with external tools. ER Flow provides an MCP Server that exposes your data model to AI — allowing it to read your current schema, create tables, add columns, set up relationships, and more.

Step 1: Get your data model UUID

Every data model in ER Flow has a unique UUID. You'll need this to connect your IDE.

Open your data model in ER Flow, then go to Settings or look at the URL — the UUID is the long string in the path. You can also find it in the Share modal.

Copy this UUID — you'll use it as the authentication token for the MCP Server.

Step 2: Configure MCP in your IDE

The MCP Server endpoint is: https://app.erflow.io/api/mcp/{uuid}

Replace {uuid} with your data model UUID.

For Cursor: Open your Cursor settings (.cursor/mcp.json in your project root or global settings) and add:

{
  "mcpServers": {
    "erflow": {
      "url": "https://app.erflow.io/api/mcp/YOUR_UUID_HERE"
    }
  }

For Windsurf: Add the MCP server in your Windsurf configuration with the same URL.

The UUID in the URL acts as your authentication token — the server uses it to identify which data model to operate on.

Step 3: Test the connection

Once configured, your AI assistant should be able to discover the ER Flow tools. Try asking it:

"Show me the current database schema"

The AI will call the get-data-model-dbml tool, which returns your entire schema in DBML format. If this works, the connection is successful.

Step 4: Use AI to modify your schema

Now you can describe schema changes in natural language:

• "Add a comments table with id, post_id, user_id, body, and created_at"
• "Add a unique index on the email column of the users table"
• "Create a foreign key from comments.post_id to posts.id with cascade delete"
• "Add a status column to the orders table with type enum and default 'pending'"

The AI translates your instructions into MCP tool calls, and the changes appear in your ER Flow diagram in real-time.

Step 5: Understand available tools

The ER Flow MCP Server provides 25+ tools organized by entity type:

Tables: create-table, update-table, rename-table, delete-table Columns: create-column, update-column, rename-column, delete-column Indexes: create-index, update-index, rename-index, delete-index Foreign Keys: create-foreign-key, update-foreign-key, delete-foreign-key Primary Keys: set-primary-key Views: create-view, update-view, delete-view Triggers: create-trigger, update-trigger, delete-trigger Procedures: create-procedure, update-procedure, delete-procedure Batch: batch-operations — execute multiple operations in a single call Read: get-data-model-dbml — get the current schema in DBML format

The AI should always call get-data-model-dbml first to understand the current schema before making changes. The batch-operations tool is especially powerful — it lets the AI create an entire schema (tables, columns, FKs, indexes) in a single request.