We are excited to announce the availability of Snowflake’s managed Model Context Protocol (MCP) servers in public preview, giving AI agents an open-standards-based interface to connect with AI-ready data in Snowflake. Snowflake’s managed MCP servers eliminate integration complexity and management overhead. Customers can connect their Snowflake data with a variety of agentic applications from providers such as Anthropic, CrewAI and Cursor through MCP connectors to create context-rich AI agents and apps. Customers can also include data from partners such as The Washington Post, MSCI, NASDAQ and The Associated Press in their MCP servers.
Customers can now create a managed MCP server alongside their Snowflake data, allowing them to seamlessly retrieve insights from both structured and unstructured data, all while remaining within Snowflake’s secure governance boundary. This approach simplifies the application architecture because the Snowflake-managed MCP server allows AI agents to securely retrieve data from Snowflake accounts without needing to deploy separate infrastructure or build custom integrations. As a result, enterprises can expedite the delivery of generative AI applications on their Snowflake data with richer insights on a standards-based, secure and robust governance model.
This capability brings several key benefits to customers:
Simplified interoperability with the broader agentic AI ecosystem, including agentic platforms such as Anthropic, CrewAI, Cursor, Salesforce and IDE plugins.
Standards-based interface for agents to discover and invoke tools and retrieve structured and unstructured data.
Consistent governance across enterprise data, AI tools and now the MCP server, all within the Snowflake secure perimeter.
Comprehensive authentication with Snowflake’s built-in OAuth service to enable OAuth-based authentication for MCP integrations.
Trusted data from top content providers with proper attribution via Snowflake Cortex Knowledge Extensions, enabling domain-specific and context-aware insights.
With Snowflake-managed MCP servers, agents can be easily configured to interoperate without custom integrations or disparate protocols. Also, developers can streamline governance and authentication across enterprise data and AI applications.
The Snowflake-managed MCP server allows AI agents to securely retrieve data from Snowflake accounts without needing to deploy separate infrastructure. MCP clients discover and invoke tools and retrieve data required for the application. At launch, the Snowflake MCP server includes Snowflake Cortex Analyst and Snowflake Cortex Search as tools on the standards-based interface. Cortex Analyst translates natural language requests into SQL queries that run against governed data to give insights into structured data. Cortex Search enables semantic search and retrieval from unstructured documents stored in or indexed by Snowflake. Customers can also retrieve data from Cortex Knowledge Extensions in Snowflake Marketplace to get licensed content from top publishers such as The Associated Press or The Washington Post. In the near future, we will support Cortex Agents in the MCP server so that remote applications can invoke agents as tools. Customers can define the tools in different database schemas inline with their current policies and access controls. Further, they can have multiple MCP servers in the account, scoped to a specific use case. Such flexible configuration allows customers to incorporate MCP servers into the Snowflake AI Data Cloud without significant changes to their governance model and deliver a high-accuracy and performant agentic experience.
By providing managed MCP servers based on open source, community-based standards, Snowflake enables customers to adopt MCP on their own infrastructure, extending choice while preserving security.
Agents can reason and help solve problems dynamically and are transforming the application architecture. Instead of rigid API contracts and restrictive UIs, agents allow for flexible natural language experience, semantic interfaces with appropriate tool use. This evolution from rigid microservices to an agentic architecture will allow reimagined experiences and new applications. But the success of these applications depends on the quality of the data they can access. Agents need easy access to quality data from external systems for accurate context. MCP enables this access using an open-standard protocol that allows agents and external systems to communicate. For enterprises, this means AI agents can be deployed faster, connected to more systems and governed consistently across the stack.
MCP is built on host-client-server architecture. Hosts are AI applications such as Claude Desktop that provide the environment for running agents. Clients are components inside those hosts that maintain direct connections to servers. Lastly, servers surface tools and resources for the agent to use. Tools are executable functions such as querying a database or performing a task.
This design creates a predictable, open interface for connecting AI agents to various systems of record. Instead of bestowed connectors, AI agents discover available tools via standard endpoints, invoke them with structured inputs and receive results in a consistent format. MCP makes agents plug-and-play in enterprise environments, simplifying integration and improving reliability.
By embedding an MCP server directly in Snowflake, customers can easily connect AI agents to their governed enterprise data. Several benefits stand out:
Governance by design: Enforce the same trusted governance policies, from role-based access to masking, for the MCP server as you do for your data.
Reduced integration effort: With the MCP server, integration happens once. Any compatible agent can then connect without new development, accelerating adoption and reducing maintenance costs.
Extensible framework: Provide agents with access to structured data and unstructured documents. You can refine the tools to improve how agents interact with your data.
Together, these benefits make Snowflake’s MCP server a powerful enabler for enterprises seeking to deploy AI agents while keeping security, governance and trust front and center.
Snowflake’s MCP server implements the open MCP specification as a server that surfaces available tools. Customers create an MCP server object and specify the tools and metadata in the server configuration. The server does not require additional compute or incur separate charges. The server object is managed with the same Snowflake role-based access controls (RBACs), ensuring that the same user and group access controls, masking and policies apply. The Snowflake-managed MCP server supports OAuth 2.0 in compliance with MCP protocol requirements.
MCP clients that connect with the server, after requisite authentication, are able to discover and invoke these tools. Tool discovery and invocation follow the standard MCP flow. Agents query the /tools/ lists endpoint to discover the tools, and the /tools/call endpoint to invoke the tools.
Discover tools with the tools/list message:
Invoke tools with the tools/call message:
The Snowflake MCP server executes those requests using Snowflake’s API and returns results. By combining standards-based interfaces with Snowflake’s governance, the Snowflake MCP server delivers a managed, enterprise-ready solution.
Claude.ai is a next-generation AI assistant built by Anthropic and trained to be safe, accurate and secure to improve productivity. Claude provides expert-level collaboration on the things you need to get done, including critical data analysis.
To add the Snowflake MCP server in Claude.ai, click on “add custom connector” and navigate to organization connectors. Next, provide the Snowflake MCP server as a custom connector.
Once the connection is established, you can interact with your Snowflake data from Claude.ai:
By connecting with Snowflake data, Claude can securely retrieve structured and unstructured data — eliminating the need to manually upload files or repeatedly provide context about your business or product. With the Snowflake MCP server, Claude brings you these insights directly, eliminating hours of manual work and letting teams focus on strategic planning instead of information gathering.
Snowflake Cortex Knowledge Extensions enable AI applications, with proprietary context and knowledge from third-party providers and publishers, together with intellectual property protection and proper attribution for content owners. Cortex Knowledge Extensions are available on Snowflake Marketplace from top-tier providers such as The Associated Press, The Washington Post, Gannett | USA TODAY Network, Stack Overflow, Packt Publishing and PubMed (published by Snowflake).
Once a Cortex Knowledge Extension has been installed in your Snowflake account, it can be added as a Cortex Search service tool in your MCP server:
The Snowflake MCP server is available in public preview today with resources to help customers get started quickly. Setting up a server involves four main steps:
Create the tools and ensure they have requisite permissions
Create the MCP server object with the tools listed in the specification
Set up authentication with the security integration and client secrets for the client
Use a client such as Claude.ai to connect with the Snowflake MCP server endpoint
You can now process natural language queries with Cortex Analyst or document retrieval via Cortex Search. Many customers begin with simple read-only use cases before expanding into workflows that include user-defined functions (UDFs) or stored procedures for approved actions.
Snowflake’s MCP server represents an important milestone in the evolving agentic architecture of AI applications built on Snowflake. It provides a governed, open-standard interface so that AI agents can easily connect AI-ready data on Snowflake. Customers can use the structured and unstructured data for richer insights and expedite the delivery of gen AI applications.
We encourage customers to start exploring today:
Deploy your first Snowflake MCP server in 20 minutes using this quickstart.
Explore available Cortex Knowledge Extensions to integrate into your AI applications.
Read about our announcement here.
With Snowflake MCP, your Snowflake AI Data Cloud becomes the foundation for trusted AI, enabling your organization to move faster, adapt more confidently and deliver AI-driven applications that make an impact.
The era of enterprise AI is here. The pace of change has never been faster, and there has never been […]
The world is expected to create 181 zettabytes of data this year — an astonishing figure that just goes to […]
We are excited to announce the availability of Snowflake’s managed Model Context Protocol (MCP) servers in public preview, giving AI […]
Our services help both the big and small organisations in the identification of security breaches and provide outclass solutions to prevent any potential harm.