Snowflake has launched Cortex Code, an AI-based coding agent that aims to extend AI assistance beyond SQL generation and conversational analytics into data and app development tasks.

Cortex Code is designed to understand enterprise data context, including schemas, governance rules, compute constraints, and production workflows, Christian Kleinerman, EVP of Product at Snowflake, said during a press briefing.

As a result, developers and data teams can use natural language to build, optimize, and deploy data pipelines, analytics, machine learning (ML) workloads, and AI agents, Kleinerman added.

Snowflake’s existing AI features, such as Cortex AISQL and Snowflake Intelligence, help users query and analyze data.

Analysts say the native coding agent can enable enterprise teams to move from experimentation to scalable data and app deployments faster, as the agent, unlike generic coding agents, understands critical contexts such as which tables are sensitive, which transformations are expensive, which pipelines are production-critical, and how analytics, ML, and agents are supposed to work together.

That contextual understanding, according to HyperFRAME Research’s practice leader of AI stack Stephanie Walter, can reduce the manual effort required to move from an experimental idea to a reliable, governed solution that is ready to run at enterprise scale.

More so because the real risk for enterprises isn’t bad code but code that breaks governance, is expensive, or can’t scale, Walter added.

Cortex Code as a CLI tool in code editors

In order to further help enterprises scale production-grade data and AI applications, Cortex Code, apart from being made available in Snowsight, is also being made available as a command-line interface (CLI) tool via code editors like VS Code and Cursor.

The deployment as a CLI tool, according to Moor Insights and Strategy principal analyst Robert Kramer, will help developers retain their enterprise’s data context, specifically data stored in Snowflake, while still being available in the code editor of their choice, locally on their machines.

The retention of context is important at the code editor level as most development work starts there, Kramer said.

Additionally, since the enterprise context is built in at the pilot stage locally, there are fewer chances of it failing in production or when it hits the warehouse, Kramer noted.

“The same Snowflake-aware agent that helps you prototype in local development workflows can follow the work into Snowflake Workspaces, Notebooks, and production pipelines. That continuity reduces the rewrite and revalidation steps where many AI pilots stall.”

Rival data warehouse and data cloud providers are pursuing comparable strategies to embed AI assistance deeper into data and application development workflows.

While Databricks is focusing more on notebook-centric development and in-platform assistants rather than local-first workflows, Google Cloud is leaning toward analyst-driven discovery through BigQuery, Looker, and Gemini, with a stronger emphasis on in-platform experiences than on local-first, Snowflake-style continuity across environments, Kramer pointed out.

Teradata, on the other hand, is prioritizing agent orchestration, governance, and control over developer ergonomics.

“The right choice depends on whether an organization’s biggest bottleneck is experimentation, governance, or operationalizing AI at scale,” Kramer said.

Cortex Code in Snowsight, which is Snowflake’s web interface, will be made generally available soon, the company said, adding that the CLI version has been made available.