Snowflake makes Cortex AI Gateway inference, observability, and cost management GA

On Nov 3, 2026, Snowflake announced general availability of Cortex AI Gateway inference, observability, and cost management: one governed endpoint for AI clients to reach models, a trace of every request in AGENT_TRACE_TABLE, spend attribution via AI_GATEWAY_USAGE_HISTORY, shared resource budgets, and per-user quotas. This release adds trace analytics (preview) charting MCP tool calls and conversation length; Gateway Tools remain in preview.

It gives a governed, auditable chokepoint for AI traffic inside Snowflake — tracing every request and enforcing per-user spend limits — which is a governance layer that also supports AI-security investigation of agent/tool misuse, and it ties MCP tool calls to the trace.

Snowflake customers centralizing model access should enable Cortex AI Gateway to audit and cap AI usage; security/data leaders can use the trace table to attribute agent activity and enforce quotas.