Snowflake Data Warehouse
Execute zero-latency SQL against raw tables exactly when the conversation originates.
Pre-emptively Inject Row Context
Configure Named SQL queries inside the dashboard. Variables like `{customer_id}` or `{session_id}` are fully injected seamlessly on conversation load using the JWT key-pair auth, making live relational database fields query-able instantly rather than through APIs.
Intrinsically Secure Execution
The system actively sanitizes inputs. Only `SELECT` statements are permitted by the parsing engine, and an internal `LIMIT 100` constraint is intrinsically appended preventing any catastrophic full-table queries affecting the warehouse.
Queries as Callable Skills
Beyond conversation-start injection, saved queries can be exposed to the agent as skills it calls mid-conversation. Ask "what were my last five orders?" and the agent runs the named query with the customer's ID and answers from the live result set.
Multiple Warehouses, More Coming
Run several warehouse connections side by side for different teams or regions. Snowflake is supported today, with BigQuery and Redshift integrations on the roadmap.