Connect Snowflake to AI: Let Claude and Cursor Query Warehouses, Tables, and Pipelines

Why Connect Snowflake to AI?

Your Snowflake warehouse holds the data that drives business decisions. Revenue tables, event streams, customer dimensions, pipeline outputs. When you need a quick number during development or in a conversation with AI, you open Snowsight, write a query, wait for results, then copy the output back.

With an MCP server pointed at your Snowflake web console, AI agents can query your warehouse directly. Get row counts, check table schemas, pull sample data, and verify pipeline outputs without leaving your editor.

What AI Agents Can Do With Snowflake Access

Once connected, Claude, Cursor, or Codex can:

• Query tables — "What's the row count for the events table this month?"
• Inspect schemas — "What columns does the dim_customers table have?"
• Check pipelines — "When did the nightly ETL last run successfully?"
• Sample data — "Show me 5 rows from the orders fact table"
• Monitor usage — "How many credits have we consumed this week?"
• Verify transforms — "What does the output of the revenue model look like?"

How It Works

1. Open DataFaucet and enter your Snowflake Snowsight URL
2. Browse your account — click through databases, schemas, tables, query history
3. DataFaucet captures the read operations as MCP tools
4. Deploy — your AI agents now have structured access to query Snowflake

You're browsing Snowsight (Snowflake's web UI), not sharing credentials. DataFaucet captures HTTP interactions and replays them as tool calls. Your Snowflake credentials and service accounts stay where they are.

Example Queries

"How many rows are in the production.analytics.page_views table?"
"What columns does dim_users have and what are their types?"
"Show me the last 5 completed tasks in the task history"
"What databases exist in this account?"
"When was the last successful run of the daily_revenue pipeline?"

When This Saves Time

• dbt development — verify model outputs without switching to Snowsight
• Data validation — check row counts and null rates while writing tests
• Incident response — pull warehouse metrics alongside application logs
• Sprint planning — get table sizes and data freshness to scope work

Security Model

DataFaucet connects through Snowsight, inheriting your existing role and permissions. If your Snowflake role only has read access to certain schemas, the MCP server has the same constraints. No warehouse credentials are stored separately. Read-only by design.

Getting Started

Point DataFaucet at your Snowsight URL (app.snowflake.com/your-org/your-account). Browse through databases and schemas you want AI to access. Deploy. Your agents can now query Snowflake data as part of their normal workflow.

Create your Snowflake MCP server →