Decube says its MCP server can provide governed catalog, lineage, quality, glossary, and observability context to Snowflake AI agents. Snowflake’s own announcements for Horizon Context, CoCo, and CoWork point to a broader push to give enterprise AI systems shared, governed business context.
Decube has described an MCP server designed to expose governed data context to Snowflake AI agents, including Snowflake CoCo and CoWork.
In a Decube post titled “Snowflake CoCo and CoWork: Governed Context for Agentic AI”, the company says its MCP server can make data governance information available to Snowflake AI agents. According to Decube, that context includes catalog metadata, lineage, data quality scores, business glossary definitions, and observability signals.
The pitch is aimed at a common enterprise AI problem: agents that can query or reason over business data need more than access to tables. They also need context about what data means, where it came from, how reliable it is, and which business definitions should apply. Decube frames its MCP server as a way to provide that surrounding governance information to agents working in Snowflake environments.
Snowflake has separately outlined a related strategy in its blog post “Snowflake Horizon Context: The Governed Context Layer for AI, BI and Apps”. In that post, Snowflake introduces Horizon Context as a governed semantic foundation for collecting, enriching, and activating business context across AI agents, BI tools, and applications.
Snowflake says Horizon Context is intended to help systems understand business meaning, not just raw data structures. The company also connects the effort to CoCo context search, positioning context as a shared layer that can be used by applications and AI systems operating on Snowflake-managed data.
That framing is consistent with Decube’s focus on supplying governed metadata and operational signals to Snowflake agents. Decube’s post describes one third-party approach to feeding governance context into Snowflake’s emerging agent ecosystem, while Snowflake’s own materials describe the platform-level context layer it is building.
Snowflake also announced CoWork, which it says was formerly known as Snowflake Intelligence. In that announcement, Snowflake describes CoWork as a personal agent for knowledge workers and says Cortex Sense works across CoWork and CoCo to provide a shared context layer for agents.
Taken together, the Snowflake and Decube materials show a clear emphasis on governed context as a practical requirement for enterprise AI agents. Snowflake is presenting Horizon Context, CoCo, CoWork, and Cortex Sense as parts of its agentic AI direction. Decube is presenting its MCP server as a way to connect governance assets—such as lineage, glossary definitions, data quality, and observability—to those agent workflows.
The key issue for enterprises will be whether these context layers can make AI-assisted analysis more reliable without weakening governance controls. The sources do not provide independent performance results or customer deployment metrics for Decube’s MCP approach. They do, however, show that both Snowflake and ecosystem vendors are treating governed business context as a necessary foundation for agentic AI in data platforms.
Hero image prompt: Editorial-style digital illustration of an enterprise data cloud with connected nodes representing metadata, lineage, quality checks, glossary terms, and observability signals flowing into abstract AI assistants; clean modern composition, blue and white palette, no logos, no readable text, no user interface screenshots.
Hero image alt text: Abstract illustration of governed data context flowing from enterprise systems into AI assistants in a cloud data environment.
Decube has described an MCP server designed to expose governed data context to Snowflake AI agents, including Snowflake CoCo and CoWork.
According to Decube, that context includes catalog metadata, lineage, data quality scores, business glossary definitions, and observability signals.
The pitch is aimed at a common enterprise AI problem: agents that can query or reason over business data need more than access to tables.
Continue reading
