Use Snowflake with AI via MCP
If your reporting runs on Snowflake, you want to reach those numbers where the work happens. Snowflake introduced a managed MCP server in November 2025 and it is now generally available – Gov regions excepted. CompanyGPT uses it through the endpoint of the MCP server object you create, and your existing role and permission structure remains fully in place.
What is Snowflake?
Snowflake is a cloud data platform used primarily as a data warehouse: data from operational systems is consolidated, modeled and made available for analysis. Compute and storage scale independently, and the platform runs in the cloud region the customer chooses – typically inside the EU for European organizations. Access is governed by a role-based access control model (RBAC).
As a tool for AI agents, Snowflake plays to its strength in dependable numbers: revenue, inventory, order intake or service metrics are held there in reconciled form. Cortex Analyst adds semantic models that map natural-language questions to the right tables and metric definitions – an important difference from freely generated SQL.
Connection at a Glance
- MCP server
- Official (Snowflake), managed, introduced 11/2025, now GA (not in Gov regions)
- Endpoint
- MCP server object in a schema: /api/v2/databases/{db}/schemas/{schema}/mcp-servers/{name}
- Transport
- Streamable HTTP
- Authentication
- Snowflake OAuth (standard), external OAuth or a personal access token
- Operating model
- SaaS (cloud region of your choice)
- Permissions
- RBAC remains fully effective
- Covered services
- Cortex Analyst, Cortex Search, SQL
How to Connect Snowflake to CompanyGPT
Snowflake introduced an official managed MCP server in November 2025; it is now generally available – Gov regions excepted. It covers Cortex Analyst for questions against semantic models, Cortex Search for search across documents and text, and SQL tooling.
One point matters for setup: the MCP server is created as an object inside a schema – so there is no single account URL but a dedicated endpoint of the form https://<account_url>/api/v2/databases/{db}/schemas/{schema}/mcp-servers/{name}. CompanyGPT talks to this endpoint directly via streamable HTTP.
For authentication, Snowflake OAuth (the standard route), external OAuth or a personal access token are available. Which one fits depends on whether access should be bound to a personal identity or to a dedicated technical account. In every case, RBAC stays in force – an agent can only read what the Snowflake role in use is entitled to.
In CompanyGPT the integration is then assigned by role and group. A tight setup is common: a dedicated Snowflake role is created with exactly the schemas required and used only for AI access. That also makes activity easier to trace later in the query history.
Typical Use Cases
How AI agents in CompanyGPT work with Snowflake
Sales figures without detours
Sales asks about win rates or revenue per region and gets the answer through Cortex Analyst, based on the semantic model that has been released.
Prepare reports
Controlling has anomalies in monthly figures summarized and moves the key statements straight into the reporting template.
Search documents in the warehouse
Using Cortex Search, the agent finds passages in contracts or minutes stored as text in Snowflake and relates them to the question at hand.
Frequently Asked Questions
Snowflake and AI – the most important answers
Is the Snowflake MCP server officially available?
Yes. Snowflake introduced the official managed MCP server in November 2025 and it is now generally available, Gov regions excepted. It is created as an object inside a schema of your account and addressed through that object's endpoint, so no installation of your own is required.
Do our Snowflake roles still apply?
Yes. Snowflake's role-based access control applies unchanged. An agent only sees the databases, schemas and tables the role in use is entitled to.
Which authentication methods are supported?
Snowflake OAuth as the standard route, external OAuth and a personal access token are possible. The choice depends on whether access should map to a personal identity or to a dedicated technical account.
How does this look from a data protection perspective?
CompanyGPT runs in your own cloud tenant and connects directly to your Snowflake account. No additional provider is added, and where your data is stored is still determined by the Snowflake region you chose.
Can arbitrary SQL queries be run as well?
Beyond Cortex Analyst and Cortex Search, the MCP server also exposes SQL tooling. Which of it may be used is defined through the permissions of the connected Snowflake role and the agent configuration in CompanyGPT.
Snowflake and Your AI – GDPR-Compliant with CompanyGPT
CompanyGPT is the enterprise AI assistant in your own cloud tenant: all relevant AI models, your company data and integrations like Snowflake – centrally managed, with roles and permissions. We set up the connection together with you.
Snowflake is a trademark of its respective owner (Snowflake). Its mention describes compatibility and does not imply any partnership or endorsement. Availability and scope of the connection depend on the respective vendor.