Snowflake has launched Cortex AI Gateway, a control plane designed to govern how enterprise AI agents access data, tools, and models. Announced July 28 at Black Hat, the gateway covers first-party agents such as CoWork and CoCo as well as third-party systems built on Claude Code and Cursor.
The product addresses two barriers to agentic AI adoption: security and cost. It centralizes access policies and authentication across more than 100 MCP servers, logs every agent action end to end, routes requests to approved models, and lets finance teams enforce spending limits before token bills balloon.
Snowflake built the gateway on its May acquisition of Natoma, an enterprise MCP platform, and is pairing it with integrations from 1Password, Aembit, Linx Security, Okta, SailPoint, and Saviynt that make third-party agents auditable. The company also announced agent identity controls, data exfiltration prevention, and ransomware protection for AI workloads.
Enterprise AI is moving from data interoperability to agent interoperability, and security has to be at the center of that shift, said chief security and trust officer Mayank Upadhyay. BlackRock and Thomson Reuters are already applying Snowflake’s zero-trust principles to their agent deployments, and Meltwater says the gateway makes its agents more observable and production-ready.
