Snowflake Expands CoCo Coding Agent With Cloud Agent and Automation

Snowflake announced an expansion of its CoCo coding AI agent platform at Snowflake Expedition 2026 on November 3, adding two new products: CoCo Cloud Agent and CoCo Automation. The company also introduced Snowflake DataStream and Interactive Analytics at the event, according to digitaltoday.co.kr.

CoCo Cloud Agent runs in an isolated cloud environment managed by Snowflake rather than on an engineer's laptop. Each session gets its own security container, and work continues after the engineer logs out. Engineers can limit the scope of the agent's permissions, and code changes go through existing review procedures via GitHub integration.

CoCo Automation lets engineers describe repetitive tasks, such as pipeline checks or investigating failed jobs, in natural language and have CoCo perform them on a schedule. When CoCo finds a problem, it identifies the cause and proposes a fix, with the engineer making the final decision on deploying it. Teams can also use the agent in Slack channels.

New data and analytics products

Snowflake DataStream lets companies using Apache Kafka send data directly to Snowflake without managing brokers or connectors. Minted, Nexon and Zywave are already using it. Interactive Analytics removes the need for a separate serving database and returns answers to dashboards, apps and AI agents within one second under heavy load, with incoming data usable for analysis within about three seconds.

Snowflake also added features for Apache Spark users, including CoCo converting existing Spark code for Snowflake. Snowpark Connect increased Iceberg table support to reduce the burden of code modification.