Can an agent inspect the schema before writing a query?

It works through SYSTEM_EXECUTE_SQL, running statements (including catalog queries) governed by the caller's role, and through CORTEX_ANALYST_MESSAGE, which generates SQL against a defined semantic model. Schema awareness comes from those paths rather than a separate introspection tool.

Why is Snowflake second and not first for SQL analytics?

ClickHouse, the top pick, is a columnar real-time engine that often wins on raw aggregation speed. Snowflake is the default when your analytics already live in Snowflake, where governed SQL and the Cortex Analyst semantic model matter more than peak query speed.

Snowflake for SQL analytics

Pick 2 of 4 for SQL analyticsOfficialSnowflake

For SQL analytics, Snowflake is our second pick of four, and it is the default the moment your analytics stack centers on Snowflake. Its managed server lets an agent run governed SQL against the warehouse and ask questions through a semantic model, which turns a vague business question into a query and an answer without a human in the loop.

It sits just behind ClickHouse, the columnar real-time engine that tends to win on raw analytical speed, and ahead of BigQuery and DBHub (Postgres). The rank is about where the data lives: if it lives in Snowflake, this is the server you reach for first.

How Snowflake fits

The tools that carry analytical work are SYSTEM_EXECUTE_SQL, which runs a statement directly and can be restricted to read-only and governed by the caller's role, and CORTEX_ANALYST_MESSAGE, which answers a natural-language question by generating governed SQL against a Cortex Analyst semantic model or view. Together they cover both the hand-written query and the asked-in-English one. CORTEX_SEARCH_SERVICE_QUERY adds hybrid search over unstructured content when an answer spans documents, and CORTEX_AGENT_RUN strings multiple steps together for a more involved question.

The honest comparison: ClickHouse, our top pick here, is built for fast aggregation over large tables and often answers a heavy analytical query quicker. BigQuery is the serverless option for Google Cloud teams, and DBHub (Postgres) covers a general-purpose relational database doing analytics double duty. Snowflake's edge is the semantic-model path and governed execution, so an agent writes correct SQL and respects policy. Reach for it when the warehouse is Snowflake; reach for ClickHouse when the priority is raw query speed on a columnar engine.

Tools you would use

Tool	What it does
CORTEX_ANALYST_MESSAGE	Answers natural-language questions over structured data by generating governed SQL against a Cortex Analyst semantic model or view.
CORTEX_SEARCH_SERVICE_QUERY	Searches unstructured content (documents, text) through a Cortex Search service using hybrid vector and keyword retrieval, ideal for RAG.
SYSTEM_EXECUTE_SQL	Executes a SQL statement directly against the account; can be restricted to read-only and is governed by the caller's role and policies.
CORTEX_AGENT_RUN	Invokes a Cortex Agent to orchestrate multi-step reasoning across Cortex Analyst, Cortex Search, and other configured tools.
GENERIC	Wraps a user-defined function or stored procedure as an MCP tool so agents can call custom Snowflake logic by name.

Full Snowflake setup and config →

FAQ

Can an agent inspect the schema before writing a query?: It works through SYSTEM_EXECUTE_SQL, running statements (including catalog queries) governed by the caller's role, and through CORTEX_ANALYST_MESSAGE, which generates SQL against a defined semantic model. Schema awareness comes from those paths rather than a separate introspection tool.
Why is Snowflake second and not first for SQL analytics?: ClickHouse, the top pick, is a columnar real-time engine that often wins on raw aggregation speed. Snowflake is the default when your analytics already live in Snowflake, where governed SQL and the Cortex Analyst semantic model matter more than peak query speed.