By Source · Snowflake

Snowflake + Connect AI

Query Snowflake warehouses from any AI agent with standard SQL. Connect AI handles role context, warehouse selection, and semi-structured VARIANT data.

Use with Claude  ·  Use with OpenAI

Common objects
Table View Warehouse Role Stage Sequence + stored procedures, file formats 
-- Ask Claude Code (or any MCP-aware agent):
--
-- "In Snowflake, show me the top 10 customers by total
-- revenue in the last 90 days. Include their region."
--
-- Claude generates and runs this SQL via the queryData MCP tool.
-- Connect AI applies the configured role and warehouse automatically.

SELECT
  c.region,
  c.customer_name,
  SUM(o.amount) AS revenue_90d
FROM [Snowflake].[ANALYTICS].[ORDERS] o
JOIN [Snowflake].[ANALYTICS].[CUSTOMERS] c
  ON o.customer_id = c.customer_id
WHERE o.order_date >= DATEADD('day', -90, CURRENT_DATE())
GROUP BY c.region, c.customer_name
ORDER BY revenue_90d DESC
LIMIT 10

Tutorials by AI surface

Working integrations for Snowflake + every supported AI surface.

Snowflake + Claude Agent SDK Warehouse analytics, ad-hoc SQL exploration, and metric questions answered via a Python agent.
Snowflake + OpenAI SDK OpenAI Python SDK assistant for warehouse queries, KPI dashboards, and natural-language drill-down.
Snowflake + CrewAI Multi-agent crew for data-driven research crews that pull live warehouse facts into briefs.

Known quirks

Case sensitivity

Snowflake folds unquoted identifiers to uppercase. Querying SELECT * FROM orders may return zero rows if the table was created as "Orders". Connect AI passes identifiers through quoted. Use the canonical case from getTables to avoid surprises.

Warehouse cost

Every query consumes credits on the active warehouse. Auto-suspend usually fires after a few minutes of inactivity, but a chatty agent making rapid tool calls can keep the warehouse hot. Size the warehouse and tune auto-suspend before pointing a production agent at it.

VARIANT and semi-structured data

JSON columns stored as VARIANT need : path or FLATTEN() to query nested fields. Connect AI exposes the columns but won't auto-flatten arrays. Tell the agent the schema shape in its system prompt or pre-flatten in a view.

Role context

The role on the Connect AI connection determines what tables are visible. If an agent reports "table not found" but the table exists, check the connection's role, not the user's default role in Snowflake.

Existing tutorials

Long-form recipes from the CData Knowledge Base.

View tutorials
Connect AI documentation

Full API reference, authentication guides, and configuration details.

View docs