Skip to main content
    Kriv AI

    Claude connectors

    Claude and Snowflake: what the connector does and what a data team should settle first

    A plain guide for data, analytics and IT leads. Every fact about the connector comes from Anthropic's directory page and Snowflake's documentation, read on 10 November 2026.

    The Claude Snowflake integration is a connector made by Snowflake and listed in Anthropic's connector directory with an Anthropic verified badge. The directory page says it can retrieve both structured and unstructured data and lists Cortex Agents and Cortex Search as its tools. Snowflake's documentation says access uses Snowflake OAuth by default and role-based access control, and that access to the server does not give access to the tools. Neither page states which Claude plans include it. This page covers what it is and what to decide first.

    plain answer

    What is the Snowflake connector for Claude?

    The Claude Snowflake integration is a connector made by Snowflake. Anthropic's directory describes it as a way to "retrieve both structured and unstructured data" and quotes Snowflake: "You can use natural language to execute SQL queries, making analytics accessible to all including business users." It lists Cortex Agents and Cortex Search as tools, is filed under Data & analytics, carries an Anthropic verified badge and says it was added in November 2025. No beta label appears on the page.

    Snowflake's documentation describes the underlying server as one that "lets AI agents securely retrieve data from Snowflake accounts" and says it can be set up to serve Cortex Analyst, Cortex Search and Cortex Agents as tools.

    Source: Anthropic connector directory, Snowflake (https://claude.com/marketplace/connectors/snowflake), and Snowflake documentation (https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-agents-mcp), read 10 November 2026.

    plain answer

    Who uses Snowflake with Claude?

    The directory page speaks of analytics for business users as well as data teams. Our reading, not a claim from Snowflake or Anthropic: analysts and finance or operations managers who want answers from governed data without writing SQL are the people who would ask the questions, and data engineers or platform admins are the people who set the server up, because Snowflake's documentation requires specific privileges to create it.

    Source: Anthropic connector directory, Snowflake (https://claude.com/marketplace/connectors/snowflake), and Snowflake documentation, read 10 November 2026.

    how to

    How do you connect Claude to Snowflake?

    Anthropic's directory page shows an Add to Claude button and marks sign-in as required. It gives no further steps. Snowflake's documentation covers the Snowflake side: a server has to be created in the account, which needs the CREATE privilege on MCP SERVER, and a person who connects needs USAGE on that server. Each user needs a default warehouse set.

    For sign-in, Snowflake says it uses "Snowflake OAuth by default, with optional External OAuth". Your Snowflake admin owns these steps; Snowflake's documentation is the one to follow.

    Source: Snowflake documentation, Snowflake-managed MCP server (https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-agents-mcp), read 10 November 2026.

    security

    What should a team decide before turning it on?

    Snowflake says the server has "role-based access control (RBAC) for the MCP server and tools" and that "Access to the MCP Server does not give access to the tools." So permission to connect is one decision and permission to use each tool is another.

    Four questions are ours, not Snowflake's. Which roles should Claude ever run as? Which databases and tables are in scope? Who reviews an answer before it feeds a decision or a report? Who reads the usage afterwards? Settle them with your Snowflake admin before the first connection.

    Source for the access statements: Snowflake documentation (https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-agents-mcp), read 10 November 2026. The four questions are Kriv's own suggestions.

    kriv role

    Where Kriv helps

    Kriv AI designs, builds and governs AI agents for business on the automation tools a company already runs, with explicit scopes, logged actions and a person approving sensitive decisions. It also runs MLOps governance work. Teams that want Claude working on their data can start with the agentic AI and automation service and agree the access rules above.

    Kriv did not build the Snowflake connector, does not speak for Snowflake, and this page describes no Kriv build with it.

    limits

    What this page could not confirm

    The pages we read do not say which Claude plans include the connector or what it costs. Snowflake's documentation states no preview, beta or general availability status, so this page states none. We read the claude.ai directory listing and Snowflake's server documentation; we did not read anything about Claude Code with Snowflake, and we did not test the connector. Read 10 November 2026.

    Straight answers

    Frequently asked questions about Claude and Snowflake: what the connector does and what a data team should settle first

    Is the Snowflake connector made by Anthropic?

    No. Anthropic's directory lists Snowflake as the maker and shows an Anthropic verified badge (read 10 November 2026).

    What tools does the directory list?

    Cortex Agents and Cortex Search (read 10 November 2026).

    How does sign-in work?

    The directory page marks sign-in as required. Snowflake's documentation says Snowflake OAuth by default, with optional External OAuth (read 10 November 2026).

    Does access to the server give access to the data tools?

    Snowflake says it does not: "Access to the MCP Server does not give access to the tools" (read 10 November 2026).

    Does this page say Claude Code works with Snowflake?

    No. The sources we read cover the claude.ai directory listing and Snowflake's server documentation, not Claude Code (read 10 November 2026).

    Talk to the team that would do the work

    Bring your requirements to a working session with the person who'll actually deliver.

    Book a Discovery Call