Stop Moving Data. Start Talking To It.

The dream of the "open lakehouse" has always been about freedom — the freedom to own your data and access it from your preferred engines and tools. But even as Apache Iceberg™ makes this a reality, this freedom comes with a hidden "metadata tax."

If you wanted to use Iceberg tables managed by an external catalog (such as Polaris, Microsoft OneLake, AWS Glue or Unity), you were signing up for a weekend of manual plumbing — juggling IAM roles, mapping schemas and individually creating those Iceberg tables to keep the various catalogs in sync.

Today, we are making "open" feel "native." With catalog-linked databases (CLD) and Cortex Code CLI, the data stays exactly where it is; the setup takes seconds; and the insights are just a short conversation away.

Catalog-linked databases: The bridge to your data

A catalog-linked database is more than a simple "connection" — it's a dynamic, live and federated link. It allows Snowflake to point directly at an external Iceberg REST catalog and prioritize it to deliver:

Zero-copy interoperability: Your data lives in your S3 or Azure bucket. You own the bits. Snowflake simply "links" to the metadata.
Live metadata sync: When a Spark job adds a new column or a Flink stream appends rows to your Iceberg table, it shows up in Snowflake automatically without any additional work.
Universal governance: This is the "secret sauce." Even though it is managed by a different catalog, you can still apply Snowflake Horizon policies — like row-level security or data masking — directly to the link, extending your existing governance to virtually your entire data estate without complicated processes.

The "easy button": Agentic infrastructure with Cortex Code CLI

The most tedious part of the open lakehouse has always been the "plumbing" — setting up external volumes for storage and catalog integrations for metadata. Traditionally, this required a 20-step manual of CLI commands and JSON policy editing.

We’ve replaced that manual with Cortex Code Skills. This goes way beyond acting as a chatbot; with Skills, Cortex Code CLI becomes an agentic assistant that understands the complex dependencies of your infrastructure.

The "Cortex Code dividend": Guided, not blind

Cortex Code CLI handles the heavy lifting on the Snowflake side while acting as a guide for your cloud provider’s setup. It works with you, rather than doing things in the dark.

From days to 60 seconds: For example, tell the CLI: I want to connect my external Iceberg catalog to a new database called 'LAKE_ANALYTICS'.

In the background, Cortex Code kicks off an agentic workflow:

Snowflake orchestration: It automatically generates and prepares the Snowflake side objects — the external volume, the catalog integration and the CLD — mapping out exactly what is needed.
Guided cloud setup: Since cloud providers require secure human intervention, Cortex Code generates the exact IAM policy and trust relationship snippets you need. It tells you precisely where to paste them in your AWS or Azure console to bridge the gap.
Handshake validation: Once you’ve updated your cloud settings, Cortex Code "tests the wires." It validates the connection in real-time to help ensure the metadata is flowing before you ever run your first query.

The payoff: Talk to all your data, wherever it lives

The real magic of a catalog-linked database isn't the setup — it’s the fact that once the connection is live, your lakehouse is immediately accessible for AI-powered insights.

Because Snowflake understands the context of your external Iceberg tables, you can use Cortex AI agents to bypass SQL entirely. Since the metadata is linked, Snowflake understands the schema, the column comments and the business logic without you having to explain it. Cortex Code CLI goes one step further. It can automatically build semantic views and agents for you for even deeper, richer analytics and recommendations.

The data didn't move. The security didn't break. And your business users didn't have to wait for a data engineer to write a pipeline. And, with Snowflake Intelligence, they can get insights in natural language.

Get started in 5 minutes or less

You don't need a three-month migration plan to start using CLDs. You just need a prompt.

Step 1: Install and link

Install the Cortex Code CLI and tell Cortex: Link my external catalog. Complete the cloud handshake using the snippets Cortex Code provides.

Step 2: Instant business insights (trends and outliers)

Once verified, your schemas populate automatically. You can immediately use Snowflake’s intelligence to learn about the business data you just connected. No complex modeling required — just ask:

"What were our top-selling product categories in the external sales lake last month?"
"Identify any unusual spikes in customer returns from our S3-hosted regional data."

Step 3: Agentic intelligence (recommendations and action)

Ready to turn data into a plan? This is where you activate Snowflake Intelligence on the data you just connected to build a specialized agent. By telling Cortex, "I want to use Snowflake Intelligence on the external catalog I just connected," you move from "What happened?" to "What should I do?"

The agent becomes a strategic partner that uses your entire lakehouse data to:

Pinpoint focus: "Based on the inventory trends in the Iceberg lake, which regions are at the highest risk of a stockout next week?"
Make recommendations: "Which customer segment should we target for our upcoming spring promotion based on the purchasing patterns we just connected?"
Formulate plans: "Create a step-by-step action plan to improve our delivery times based on the shipping logs in our external tables."

The future is linked

We are moving away from the era of "data warehouses" and "data lakes" and into the era of the connected data estate. By combining the openness and interoperability of Iceberg with the simplicity of Cortex Code Skills, we’re making it possible to manage your entire data estate through a single, simplified and secure conversational control plane.

The "single pane of glass" isn't a dashboard anymore. It’s a prompt. It’s time to stop moving your data and start talking to it.

Ready to see Cortex Code CLI in action? Download the Cortex Code CLI and link your first catalog today.