• SCHEMA
  • LOGIC
  • SOURCE
  • HUMAN

How to audit SQL data models with your AI

Connect Hightouch to your AI client to pull warehouse schemas and SQL definitions into your chat. You can then ask your agent to find inconsistencies or broken logic.

Ask AI about this page

Short answer

How can I audit SQL data models with my AI?

Your AI uses Hightouch to pull your warehouse source schemas and SQL definitions. This lets you check for schema drift or logic errors without leaving your chat interface. You'll get a clear view of how your data is structured and defined.

Where every issue lands.

Audit outcomes

The AI organizes its findings into these specific categories.

SCHEMA

Schema drift detection

The AI compares your current warehouse state against expected definitions. You'll see exactly where columns or types have changed.

LOGIC

SQL definition review

Your agent inspects the SQL code used in your models. It flags potential errors in joins, filters, or aggregations.

SOURCE

Source data validation

The AI checks the raw source schemas to ensure the data being pulled matches what your models expect.

HUMAN

Manual fixes required

The AI identifies complex logic errors it can't resolve alone. You'll receive a list of specific SQL blocks that need your direct attention.

The workflow

What your AI does when the data arrives.

The AI acts as a bridge between your warehouse and your chat window.

  1. Map the warehouse

    The AI identifies all available data sources in your warehouse to establish a baseline.

    list_sources

  2. Identify models

    The agent pulls a list of all existing data models to see what needs auditing.

    list_models

  3. Inspect definitions

    Your agent reads the specific SQL logic for each model to find bugs.

    list_models

  4. Verify schemas

    The AI cross-references model requirements against the actual source structures.

    list_sources

Try it

Copy these to start.

Use these prompts to begin your audit.

Starting points

These are opening lines. Swap out the specific model or source names to match your warehouse.

Hightouch Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_Ol0VUkS1D1nD6Pbi85YceDBC5tsYdocr5u8J49aY/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — Hightouch capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "hightouch-reverse-etl-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_Ol0VUkS1D1nD6Pbi85YceDBC5tsYdocr5u8J49aY/mcp"
    }
  }
}
Copy into chat04

  • List all my data models and tell me which ones use the 'sales_events' source.

  • Check the SQL definition for the 'monthly_active_users' model for any broken joins.

  • Compare the schema of my 'customer_dim' source with the models that depend on it.

  • Find any models that are referencing columns that no longer exist in my warehouse sources.

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect Hightouch once, then ask.

Just link your Hightouch account. Your credentials stay encrypted, and you can immediately start asking your AI about your data models.

Connect Hightouch to your AI

FAQ

How this task behaves

  • 01

    Can the AI change my SQL code?

    No. The AI can only read and inspect your SQL definitions. You must manually apply any changes to your warehouse.

  • 02

    Can the AI delete my data sources?

    No. The tools only allow the AI to list and read information. It has no permission to delete or modify anything.

  • 03

    How does the AI know my schema changed?

    The AI uses the list_sources tool to see the current state of your warehouse and compares it to the definitions it reads.

  • 04

    Does the AI see my actual row-level data?

    No. The AI only sees the metadata, schemas, and SQL definitions provided by the tools.

  • 05

    Can I audit models from multiple warehouses?

    Yes, as long as those sources are connected through Hightouch and visible to the AI.

  • More questions about Hightouch? The Connector page answers them. See everything the Hightouch Connector can do

Connect Hightouch to Claude, Cursor, ChatGPT & more