• VALIDATION
  • LOGIC
  • DISCOVERY
  • HUMAN

How to audit dataset schemas and recipes with your AI

Connect your agent to Dataiku DSS to inspect your data pipelines. You can then query specific configurations without clicking through the DSS interface.

Ask AI about this page

Short answer

How can I audit my Dataiku DSS schemas and recipes using AI?

Your agent pulls dataset schemas and recipe configurations directly from Dataiku DSS. This lets you verify data structures and transformation logic through natural language. You'll get a clear view of your pipeline's current state.

Audit outcomes

Where your findings land.

The agent translates raw Dataiku metadata into clear answers about your data flow.

VALIDATION

Schema drift detection

The AI compares current dataset structures against your expectations. You'll see exactly where columns or types have changed.

LOGIC

Recipe breakdown

Your agent reads the transformation steps in your recipes. It explains the math or logic used to create new datasets.

DISCOVERY

Pipeline mapping

The AI lists all available recipes in your project. You can quickly map out how data flows from one stage to the next.

HUMAN

Manual fixes

The AI identifies errors but cannot change your DSS code. You'll need to go into the Dataiku UI to apply any corrections.

The workflow

What your AI does when the audit starts.

The agent follows a sequence of inspection steps to build a complete picture of your data environment.

  1. Map the project

    The agent identifies all the transformation steps currently active in your Dataiku project.

    list_recipes

  2. Inspect transformations

    It pulls the specific configuration and logic for a single recipe to understand how it processes data.

    get_recipe

  3. Check data types

    The agent looks at the actual columns and data types within a specific dataset.

    dataset_schema

  4. Analyze inconsistencies

    By combining the schema and recipe info, the agent flags where logic might fail due to unexpected data changes.

    dataset_schemaget_recipe

Try it

Copy these to start.

Use these prompts to begin your audit.

Starting points

Replace the bracketed text with your actual Dataiku dataset or recipe names.

Dataiku Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_omB2MsN4Y1V0ZX0dD6ibopTA1n2aAb9aFGAtDQOY/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 — Dataiku capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "dataiku-dss-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_omB2MsN4Y1V0ZX0dD6ibopTA1n2aAb9aFGAtDQOY/mcp"
    }
  }
}
Copy into chat04

  • List all the recipes in my Dataiku project so I can see the pipeline.

  • Show me the schema for the 'customer_orders' dataset.

  • Explain the transformation logic used in the 'clean_sales_data' recipe.

  • Compare the schema of 'raw_logs' with the output of the 'process_logs' recipe.

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

Start here

Connect Dataiku DSS once, then ask.

One connection gives your agent access to your Dataiku project. Your credentials stay encrypted, and you can start auditing immediately.

Connect Dataiku DSS to your AI

FAQ

How this task behaves

  • 01

    Can the AI change my Dataiku recipes?

    No. The AI can only read schemas and recipe configurations. You must make all changes manually in Dataiku DSS.

  • 02

    Can the AI run the recipes for me?

    No. This MCP is for inspection and auditing. It does not have the capability to trigger recipe execution.

  • 03

    How does the AI see my data?

    The agent uses the provided tools to pull metadata like column names, types, and recipe logic. It does not browse your entire file system.

  • 04

    Does this work with any AI client?

    Yes, it works with any MCP-compatible client like Claude, Cursor, or Windsurf.

  • 05

    Can I audit multiple projects at once?

    The agent can inspect any project it has permission to access through the connection you set up.

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

Connect Dataiku to Claude, Cursor, ChatGPT & more