Snowflake Unveils Cortex AI Gateway for Governing Enterprise Agents
Snowflake introduced Cortex AI Gateway on July 28, 2026, as a centralized layer for controlling how first- and third-party AI agents access models, tools, MCP servers, data and enterprise systems. The company says the gateway will enter public preview soon and will add policy, activity and cost controls, giving security, engineering and finance teams one place to govern agent access and consumption.
Snowflake introduced Cortex AI Gateway on July 28, 2026, as a centralized control layer for AI agents operating across enterprise systems. The company says the product will enter public preview soon, so this is an announced capability rather than a generally available service.
One gateway for agent access
Cortex AI Gateway is designed to govern how Snowflake-built agents and third-party agents connect to models, data, applications, tools and Model Context Protocol servers. Snowflake says the gateway will centralize access policies, authentication and permissions, with support for more than 100 MCP servers.
The system is also intended to keep an end-to-end activity record showing which resources an agent used and the order of its actions. For teams running several agent platforms, that could reduce the gap between identity controls, application logs and AI-specific observability. Whether the controls work consistently across third-party systems will depend on the integrations and their eventual availability.
Cost controls and model routing
Snowflake says the gateway will attribute AI consumption to teams, agents and workloads, and will let organizations enforce spending limits. It also plans to route requests among enterprise-approved models based on quality, cost, latency and availability. Those functions address a practical problem for data and platform teams: agent workloads can cross several providers while generating usage faster than traditional monthly cost reviews can catch.
The gateway incorporates capabilities from Natoma, the enterprise MCP platform Snowflake acquired in May 2026. Snowflake also announced integrations with 1Password, Aembit, Linx Security, Okta, SailPoint and Saviynt. The availability details matter: most of those integrations are planned for private preview soon, while the Okta integration is planned for private preview in the fourth quarter of 2026.
For practitioners, the immediate task is evaluation rather than migration. Teams should compare the announced controls with their existing identity, audit-log and model-routing requirements, then validate enforcement and cost attribution once the preview is available. Snowflake has described the control plane; production evidence will need to show how reliably it governs agents outside Snowflake.
Key Points
- 1Snowflake announced Cortex AI Gateway on July 28, 2026, with public preview planned soon.
- 2The gateway is designed to centralize agent access policies, activity records, model routing and AI consumption controls.
- 3Snowflake says the product will support more than 100 MCP servers and both first- and third-party agents.
- 4Several identity integrations remain planned for private preview, so production enforcement still needs validation.
Scoring Rationale
High enterprise data and AI relevance because the announced gateway combines agent access governance, observability, model routing and cost controls, tempered by preview-stage availability and limited independent production evidence.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
