Build Faster with Cortex Code — Now More Widely Available and Ready for Bigger Tasks

Cortex Code, Snowflake’s AI coding agent, is evolving rapidly and becoming part of the day-to-day workflow for more data practitioners across every team. From seasoned engineers to less technical builders, it helps teams bring ideas to life faster, put them into production and scale them with confidence.

In today’s update, we are thrilled to share that Cortex Code is expanding its availability and reach, as well as the scale and variety of assignments it can handle. This announcement brings four major updates: Cortex Code in Snowsight is now generally available; the CLI now supports native Windows environments; Agent Teams make it easier to break large assignments into coordinated parallel work; and new agent skills standardize how Cortex Code helps build on data.

Cortex Code in Snowsight: A persistent agent for every Snowflake user

For many Snowflake users, Snowsight is where the day-to-day work takes place. Now, every Snowflake user has access to Cortex Code directly inside their secure and governed Snowflake environment. 

Designed as a persistent AI coding agent in Snowflake Workspaces, Snowflake Notebooks and other Snowsight workflows, Cortex Code in Snowsight is fully aware of where you are in the UI and what data you're working with, and it has the latest knowledge of your specific data catalog: databases, schemas, tables, semantic models and more.

That changes the day-to-day experience in a meaningful way. Instead of treating AI assistance as a separate destination, teams can use Cortex Code where the work is already happening:

• Agentic coding in Workspaces: Generate, modify, optimize and explain SQL and Python code through conversation. Preview AI-suggested changes in a diff view before applying, fix failed queries with one click, and get context-aware inline code suggestions as you type.

• Accelerate end-to-end building in Snowsight: Develop agentic ML workflows in notebooks, stand up dbt pipelines, create semantic models for Snowflake Cortex Analyst, or coordinate broader data engineering tasks — all in natural language.

• Data and documentation discovery: Search for database objects, tables and columns using plain language without knowing exact names. Get answers about Snowflake features, SQL syntax and best practices pulled from official docs and the Horizon Catalog, including tags, masking policies and lineage context.

• Agentic administration and cost management: Query governance, security and spend information conversationally — such as identifying PII-tagged tables, auditing role access, monitoring credit consumption and pinpointing high-cost warehouses or queries.

Native Windows support brings the CLI to more developer teams

Cortex Code CLI already gave developers the power to build on their enterprise data without leaving their editors such as VS Code, Cursor or the terminal. With this release, that experience extends to Windows.

To install Cortex Code CLI on Windows, issue the following command in PowerShell:

Cortex Code CLI simplifies and automates the tedious parts of end-to-end development on Snowflake and other data systems while:

• Accelerating the speed of building and deploying applications, agents, data pipelines and more

• Understanding and handling the nuances of Snowflake’s data, compute, governance and operational semantics

• Offering an ever-growing list of specialized skills to simplify your data workflows. Check out how Cortex Code CLI supports popular data systems like dbt and Apache Airflow®.

Agent Teams help Cortex Code take on bigger, multipart tasks

Agent Teams introduce a new coordination layer that transforms Cortex Code into a powerful multiagent orchestrator. Instead of tackling tasks one at a time, you can spin up independent subagents that run in parallel, each assigned a specific job and persona.

A lead agent can spin up teammates with defined roles, such as research, coding or testing, and coordinate them through a shared task list. Work can be assigned, updated and completed in parallel, with dependencies keeping tasks in the right order and outputs flowing cleanly from one agent to the next.

Users stay in control throughout. You can step into any live agent session to review progress, respond to approvals or redirect work, then return to the lead session without breaking context.

New agent skills make Cortex Code more adaptable to your team’s workflows

A skill is a packaged set of instructions, context and workflows that Cortex Code can load to perform expert data tasks on Snowflake and your data stack. 

Think of it like giving the agent a specialist playbook — when you ask about cost optimization, ML deployment or Streamlit styling, the matching skill activates and equips the agent with the right docs, best practices and step-by-step guidance so it doesn't have to figure things out from scratch. Skills keep responses accurate and domain-aware without bloating every conversation with everything the agent knows. Here are some of the latest skills we have introduced:

• cortex-code-guide: New to Cortex Code? This skill walks you through everything — from available commands and session management (resume, fork, rewind) to agent configuration, keyboard shortcuts, MCP integrations and the #table syntax for referencing Snowflake objects right in your prompts. Think of it as your interactive cheat sheet for getting productive fast.

• developing-with-streamlit: Build, debug and style Streamlit apps without leaving your flow. Whether you're wiring up a quick dashboard or fine-tuning CSS and themes, this skill handles the scaffolding, component patterns and deployment so you can focus on what your app actually does.

• openflow: Move data where it needs to go. Snowflake Openflow taps into Snowflake's NiFi-based engine to help you deploy connectors, configure replication pipelines and build transformation flows — all without spinning up separate infrastructure.

• cost-intelligence: Know exactly where your Snowflake spend is going. This skill surfaces credit consumption, warehouse costs, storage trends and budget alerts — then helps you set up resource monitors and catch anomalies before they become surprises on your bill.

• machine-learning: From training to registry and inference, this skill covers the full ML lifecycle on Snowflake. Train models at scale, log them to the Snowflake Model Registry, deploy to warehouse or container services, and monitor drift — all without stitching together a dozen external tools.

• cortex-ai-functions: Classify text, extract entities, run sentiment analysis, summarize documents, translate content and more — directly in SQL. This skill helps you pick the right Snowflake Cortex AI function for your use case and wire it into your data pipeline with minimal code.

Get started today

Open Snowsight and start a conversation — with zero setup. Or grab the CLI (now with Windows support) and plug Cortex Code into your favorite editor or the terminal you prefer.

Build faster with Cortex Code. 

Try Cortex Code in Snowsight or start a Snowflake trial. 

Learn more and read our documentation.