April 28, 2026

Leverage Sigma Skills Directly in Snowflake Cortex Code

April 28, 2026

Last week, we shared how the Sigma MCP Server brings live, governed data directly into AI chat assistants like Claude and ChatGPT. Today, we're bringing that same philosophy to Snowflake Cortex Code so data engineers can build, manage, and validate their full Sigma workflow without ever leaving their Snowflake environment.

Hi, all. I'm Jordan, a product manager here at Sigma. And today, I'm gonna show off using Cortex Code with Sigma computing skills. A quick background, Cortex Code is essentially Snowflake's coding agent. We're shipping a set of Cortex Code skills for Sigma, so now when data engineers are doing work in Snowflake, they can manage everything Sigma in one workflow. So you can see here that I've got the Sigma plug in loaded. It gives me skills for everything like authentication, creating a data model, updating a data model, building workbooks, all the stuff you'd normally have to do in the Sigma UI. So let's say I'm a data engineer. I just added a new orders table to Snowflake. I can see that here by saying what tables are in my default schema. I want to create a data model on top of the orders table, using the right columns, key metrics, and a relationship to my customers table. So I just describe what I want, and watch what happens. Cocoa will authenticate with the Sigma API, discover the actual column names from Snowflake, compose the full API request with the metrics and relationships, validate it against the API, and submit it. And then I've got a live data model within Sigma. I can go and open this to see it in the Sigma UI. And here I have my data model with my orders and customer tables and some key metrics. Now the real payoff building a workbook. I can just ask Coco to do it for me. So I'll add in this prompt, and it's using the, build workbook skill. We'll give Coco a second to work here. It finds the data model, reads the right reference docs, drafts the API calls, runs validation on the formulas, and then posts it to create the workbook. Great. So the work was created, I can go see it in Sigma. You can see my executive dashboard with region filter, KPIs, trends month over month, and a breakdown by different dimensions. So the thing I'm most excited about here is that data engineers never have to leave their workflow. You can add a table, get a data model, build a dashboard quickly, you just describe it, and it's built. We're really excited about this integration and working closely with the Snowflake team to make building in Sigma from Cocoa an amazing experience. Thanks for watching.

From a single interface, teams can go from raw table to governed data model to published dashboard — and keep Snowflake and Sigma in perfect sync as the data evolves.

Cortex Code is Snowflake's AI-powered coding agent — built for data engineers, developers, and architects who want to write, debug, and deploy data workflows in Snowflake. It's powerful on its own, and now the workbooks, data models, and business logic your teams have built and validated in Sigma are accessible directly from Cortex Code — giving data engineers a complete, end-to-end workflow from a Snowflake table to Sigma AI App, all without switching tools.

Create, update, and deploy Sigma assets directly from Cortex Code

At Sigma, we shipped a set of skills for Cortex Code that teach the agent how to work with Sigma's APIs. Think of skills as expert instructions: They tell the agent which API calls to make, in which order, and how to validate the result. Once installed, the Sigma plugin surfaces directly inside the Cortex Code interface with a full set of commands: setting up the CLI, creating data models, updating existing models, building workbooks (coming soon), and managing users.

Here's what that looks like end to end:

A data engineer adds a new ORDERS table to Snowflake and asks Cortex Code to build a data model on top of it — specifying the columns, a sum (revenue) metric, and a relationship to the CUSTOMERS table.
The agent then authenticates with the Sigma API, discovers the actual column names from Snowflake, composes the full spec, validates it, and submits. The result is a live Sigma data model, and the engineer never opened a browser.
When the warehouse team adds a discount column the next week, the engineer describes the update in plain language, and Cortex Code pulls the current spec, applies the change, and uses the PUT command. Every workbook sourcing that model picks up the new column automatically.
From there, building a dashboard is just as direct. A single description is enough. The data engineer can tell Cortex Code, "Build an executive sales dashboard from the Orders data model with KPIs for total revenue, order count, and average order value, a revenue trend line chart by month, and a region filter dropdown." With that prompt, the agent finds the model, maps the features, validates the formulas, and posts it.

A data engineer can generate a Sigma workbook directly from the CLI using Snowflake Cortex Code with Sigma skills.

Permissions follow the same principle: Cortex Code inherits them automatically from Sigma, so the agent sees exactly what the user would see in Sigma. There’s no new security surface and no risk of over-permissioning.

How organizations are using Sigma Skills in Snowflake Cortex Code

1. End-to-end pipeline management

A data engineer adds a new table to Snowflake. From Cortex Code, they create a Sigma data model on top of it, define the key metrics and column relationships, and build actionable workbooks — all without leaving the terminal. When the schema changes downstream, they update the model in the same interface. What used to require a browser, a UI walkthrough, and several back-and-forth messages with an analyst now happens in a single, continuous workflow.

A data engineer asks Cortex Code to build a Sigma data model on a new Snowflake ORDERS table. The agent composes the spec, validates it, and submits — no browser required.

2. Permissions management at scale

Because Cortex Code inherits permissions from Sigma's access controls, teams can manage who sees what from the same interface where they're managing the data itself. As organizations grow and data access needs evolve, the Sigma skills allow engineers to provision users and adjust permissions programmatically — keeping governance consistent without requiring a separate admin workflow.

3. Automated reporting and scheduling

Once a workbook is built, Cortex Code can wire up the delivery layer too. Data engineers can configure scheduled report exports, set up snapshot pipelines that capture and archive BI output on a recurring cadence, and push results directly into the systems where stakeholders already work: like Slack channels, email distributions, downstream data warehouses, or external apps via API. The entire pipeline, from data model to delivered report, can be defined and automated from a single interface without touching the Sigma UI.

Start building today

Data engineers already live inside Snowflake and now Sigma does too. From a single interface, teams can go from raw table to governed data model to published dashboard — and keep Snowflake and Sigma in perfect sync as the data evolves.

Looking ahead, we're adding Input Table support so teams can create custom AI applications that go beyond reporting — capturing user inputs, triggering automated workflows, and acting on live data, all without leaving Snowflake.

To learn more about Sigma's integration with Snowflake Cortex Code, request a demo or reach out to your Sigma representative. You can also meet with the Sigma team at Snowflake Summit in San Francisco.

AI/ML

Product