The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.

Snowflake has introduced new capabilities for Cortex Agents to address operational challenges in enterprise AI deployments, including orchestration, execution, and governance. These enhancements aim to simplify building, deploying, and managing AI agents at scale within Snowflake’s governance framework. Agents can now reason over enterprise data, invoke tools, and execute actions while adhering to existing access controls and audit trails. The platform handles model routing, tool invocation, and result assembly, reducing manual overhead for developers.

Snowflake CoCo, a data-native AI coding agent, is now available for public preview via the Cortex Agents REST API. CoCo enables developers to integrate coding capabilities—such as SQL execution, data transformation, and pipeline automation—into custom applications. Configured with a sandboxed runtime, CoCo allows agents to perform tasks like debugging Cortex Agent requests without leaving Snowflake’s secure environment. The managed API provisions sandboxes dynamically, streamlining workflows like root-cause analysis for agent requests.

Two new features, Skills Package and Agent Toolset, are entering public preview to streamline agent management. Skills Package lets teams define a reusable set of capabilities once and apply them across multiple agents via a single URI, ensuring consistency. Agent Toolset allows agents to reference tools from another agent’s specification, eliminating duplication and reducing maintenance overhead. These tools are designed to support shared capability layers, such as standard data workflows or integrations, across large-scale deployments.

Additional updates include Tool Search for efficient tool library management, Async Agent API for long-running tasks, and Interrupt and Resume for real-time agent steering. The Code Execution Tool provides a secure Python sandbox for data processing and report generation. These features aim to improve scalability, accuracy, and user control in enterprise AI workflows, with several capabilities entering general availability or public preview in the coming months.