Snowflake Cortex AI Gateway enters public preview

On October 23, 2026, InfoQ Chinese reported that Snowflake's Cortex AI Gateway had entered public preview. The product is positioned as a centralized control plane for enterprise AI, offering governed model inference, MCP tool governance, observability, and cost control through a single endpoint. According to the report, it is already compatible with the Chat Completions or Messages interfaces of models including OpenAI, Gemini, Claude, and DeepSeek, and existing clients can connect by changing just one base URL and one token.

The report frames the launch against the fragmentation of enterprise AI stacks. Vector databases, monitoring systems, and pipeline tuning components are linked through coding agents, embedded agents, and MCP servers that connect to enterprise systems, while the underlying mix of models changes almost every few weeks. Different teams adopt components at their own pace, which the report says can create management and governance gaps. In such a fragmented environment, few enterprises can accurately answer basic questions across layers: which models are being called, who is calling them, what actions they are performing, and what the costs are. The report argues that simply restricting employee access to AI only pushes them toward unapproved tools, further increasing risk.

According to the report, Cortex AI Gateway acts as a control node between each AI client and the enterprise systems it accesses, providing unified visibility, governance, and oversight. The report also states that a modern enterprise AI infrastructure stack must have six key elements, though the summary does not detail what they are. The product currently remains in public preview; the material does not mention a general availability date, pricing, or customer cases.