Introduction
Cortex Knowledge Extensions (CKEs) became Generally Available (GA) on August 12, 2025. This was announced in the official blog.
The Easy Button for Context-Rich AI Agents
Discover how Snowflake Intelligence and Cortex Knowledge Extensions help enterprises unlock insights from data within their organizations as well as third-party expertise.
snowflake.com
This article will provide an overview of CKEs and introduce various ways to utilize them.
Usage Example
After acquiring the CKEs for Snowflake Documentation and configuring Snowflake Intelligence (currently in Public Preview as of October 27, 2025), it can provide answers that reference the CKEs as shown above. The setup is very simple and will be explained later in this article.
What are Cortex Knowledge Extensions (CKEs)?
First, let's explore what Cortex Knowledge Extensions (CKEs), which became GA, actually are. According to the Cortex Knowledge Extensions documentation:
Cortex Knowledge Extensions (CKEs) are Cortex Search Services that can be shared on the Snowflake Marketplace or via private listings or organizational listings. They can be used in a retrieval-augmented generation (RAG) architecture to integrate licensed and proprietary content into Cortex AI applications. For example, CKEs can be used to integrate knowledge from unstructured content, such as articles, market research, books, or forum posts, into Cortex AI applications, such as chatbots and agentic systems.
As stated in the documentation, CKEs are essentially "Cortex Search" services shared via Snowflake Marketplace, private listings, or organizational listings.
CKEs listed on the Snowflake Marketplace can be found here:
Snowflake Marketplace
Various data providers have started offering CKEs, including Snowflake-provided ones like "Snowflake Documentation" and "PubMed Biomedical Research Corpus".
How CKEs Work
As described in the documentation, the mechanism is as follows:
Providers (CKE providers) upload text data to tables in their accounts and create Cortex Search services on those tables. These Cortex Search services are shared on the Snowflake Marketplace. The Cortex Search services shared on Snowflake Marketplace become Cortex Knowledge Extensions (CKEs). When using these CKEs, you can develop applications by combining them with Cortex AI SQL or Cortex Agent API.
As mentioned earlier, CKEs are essentially Cortex Search services, which means they can be used in various ways. Let's explore what forms of utilization are possible.
Use Cases
Here are some examples of how CKEs can be integrated and used.
Snowflake Intelligence
Integration with Snowflake Intelligence is straightforward. Following the procedure outlined in the Quickstart guide below, you can register the CKE's Cortex Search as an "agent" to make it immediately available in Snowflake Intelligence.
Getting Started with Snowflake Intelligence
Snowflake Intelligence is a great option when you want to make Cortex Agent, Cortex Analyst, and Cortex Search widely available within your organization.
Snowflake Cortex AI MCP Server
"Snowflake Cortex AI MCP Server" is available on Snowflake-Labs. By configuring it to use CKE's Cortex Search, you can utilize it with any MCP client (Claude Desktop, Cursor, gemini-cli, etc.).
Snowflake Cortex AI Model Context Protocol (MCP) Server
Snowflake Managed MCP Servers
Snowflake Managed MCP Servers were released as a preview on October 1, 2025. This allows you to host Cortex Analyst / Cortex Search as remote MCP Servers on Snowflake.
Introducing Snowflake Managed MCP Servers for Secure, Governed Data Agents
Snowflake Managed MCP Servers, an open-standards interface for AI agents to access your governed data without custom integrations or separate infrastructure.
snowflake.com
Setting Up for Use with Snowflake Intelligence
In this article, let's configure CKEs for use with Snowflake Intelligence.
Prerequisites
To use CKEs, the easiest way to start is with CKEs published by Snowflake itself, such as their Documentation.
After logging into your Snowflake account, access this URL and acquire it.
Configuring Cortex Agent
To use Cortex Search with Snowflake Intelligence, you need to configure a Cortex Agent. While you can set this up via the GUI in Snowsight, it's also possible to define it using SQL. For documentation on handling Cortex Agent with SQL, please refer to this page.
After acquiring Snowflake Documentation from the Snowflake Marketplace, executing the following SQL will make it available in Snowflake Intelligence.
Note that cross-region inference is enabled in the SQL, so please be aware of this.
details Cortex Agent for CKEs Setup SQL:
USE ROLE ACCOUNTADMIN;
ALTER ACCOUNT SET CORTEX_ENABLED_CROSS_REGION = 'ANY_REGION';
CREATE DATABASE IF NOT EXISTS snowflake_intelligence;
GRANT USAGE ON DATABASE snowflake_intelligence TO ROLE PUBLIC;
CREATE SCHEMA IF NOT EXISTS snowflake_intelligence.agents;
GRANT USAGE ON SCHEMA snowflake_intelligence.agents TO ROLE PUBLIC;
GRANT CREATE AGENT ON SCHEMA snowflake_intelligence.agents TO ROLE PUBLIC;
CREATE OR REPLACE AGENT SNOWFLAKE_INTELLIGENCE.AGENTS.CKE_SNOWDOC_AGENT
WITH PROFILE='{ "display_name": "Snowflake Document" }'
COMMENT=$$ This is an agent that can answer questions about Snowflake Documentation. $$
FROM SPECIFICATION $$
{
"models": {
"orchestration": ""
},
"instructions": {
"response": "# Persona\nYou are a Snowflake expert consultant. Act as an experienced professional supporting enterprise customers.\n\n# Tone & Manner\n- Always use polite language and provide accurate, concise answers.\n- Add brief explanations for technical terms when necessary, so that even beginners can understand.\n\n# Strict Rules for Response Generation\n1. Strict adherence to sources: All answers must be based on Snowflake's official documentation. Do not include personal opinions or information that cannot be verified in the documentation.\n2. Clear source attribution: At the end of your response, include a section titled [Reference Documentation] and list the URLs of referenced pages in bullet points.\n3. Formatting:\n- Enclose code such as SQL, Python, JSON, etc., in code blocks with language specification (e.g., ```
sql ...
```).\n- Organize procedures and lists using bullet points (-) or numbered lists (1.) for readability.\n- Emphasize important keywords such as function names and parameter names with `backticks`.\n4. Handling unclear questions:\n- If an answer to a question cannot be found in the official documentation, honestly respond with \"I was unable to find a direct answer from the official documentation for your question.\"\n- Then, provide related information by saying \"Here are potentially relevant documents\" or suggest next steps like \"You might get answers from other experts by asking in the Snowflake Community.\"\n5. Prohibited actions:\n- Do not answer specific amounts regarding pricing or licenses; always provide a link to the official pricing page.\n- When asked about security-related configurations or personal information, only provide general best practices and never present specific configuration values or prompt for information input.\n\n\n",
"sample_questions": [
{
"question": "Tell me how to set up Snowflake Openflow"
},
{
"question": "What is the current status of Snowflake's Iceberg support?"
},
{
"question": "What LLM models are available in Snowflake?"
},
{
"question": "Which clouds and regions can be used when using Snowflake in Japan?"
}
]
},
"tools": [
{
"tool_spec": {
"type": "cortex_search",
"name": "Search Snowflake Documents",
"description": "Cortex Search Knowledge Extensions for searching Snowflake documentation"
}
}
],
"tool_resources": {
"Search Snowflake Documents": {
"id_column": "SOURCE_URL",
"max_results": 5,
"name": "SNOWFLAKE_DOCUMENTATION.SHARED.CKE_SNOWFLAKE_DOCS_SERVICE",
"title_column": "DOCUMENT_TITLE"
}
}
}
$$;
Using with Snowflake Intelligence
After executing the above SQL, navigate to "AI & ML" > "Snowflake Intelligence" as shown below to access Snowflake Intelligence.
After launching Snowflake Intelligence, select "Snowflake Document" and ask a question to receive results.
Conclusion
CKE is a service that allows you to easily acquire and utilize knowledge bases provided on the Snowflake Marketplace. When utilizing created Cortex Search services, you can choose different approaches based on your use case: Streamlit for building custom UIs, Snowflake Intelligence for deploying to various users within your organization, or MCP Server for data scientists, data engineers, and application developers. We hope you'll utilize CKE and Cortex Search according to your specific use cases.
Promotions
Snowflake What's New Updates (by tsubasa)
tsubasa shares Snowflake What's New updates on X. Please follow to stay updated with the latest information.
Japanese Version
Snowflake What's New Bot (Japanese Version)
English Version
Snowflake What's New Bot (English Version)
Japanese version
https://zenn.dev/tatsu_tech/articles/2695582e0acb9a
Update History
October 27, 2025: New post
Top comments (0)