Expose governed analytics
Managed servers can publish Cortex Analyst, Cortex Search, Cortex Agents, SQL execution and custom tools.
Documentation ↗MCP server / data & analytics / source-linked
Snowflake MCP is an enterprise data plane, not a small local adapter. It can expose Cortex Analyst, Cortex Search, Cortex Agents, SQL and custom tools while inheriting Snowflake roles. Compare it on governance and tool boundaries as much as on query capability.
Identify the implementation and the job it supports.
Snowflake MCP is an enterprise data plane, not a small local adapter. It can expose Cortex Analyst, Cortex Search, Cortex Agents, SQL and custom tools while inheriting Snowflake roles. Compare it on governance and tool boundaries as much as on query capability.
The documented interface, authentication and installation path.
Follow Snowflake's quickstart to create a managed MCP server and bind it to permitted schemas/tools. Set an MCP access role and default warehouse, then connect a client through the protected resource metadata flow.
Setup reference ↗Concrete capabilities with maintainer sources.
Managed servers can publish Cortex Analyst, Cortex Search, Cortex Agents, SQL execution and custom tools.
Documentation ↗Role-based permissions govern server and tool access; Snowflake or external OAuth can authenticate clients.
Documentation ↗Snowflake documents dedicated access roles, default roles and disabling broad secondary roles for MCP sessions.
Documentation ↗What was checked and what was not tested.
Public maintainer documentation and repositories. This is not a hands-on product review.
Setup examples are documented examples. Suggested trials and comparison advice are editorial guidance.
Sources support the adjacent factual statements. A source review does not establish uptime, performance, security or current account eligibility. Read the editorial policy.
A bounded evaluation you can reproduce in your own environment.
Suggested evaluation, not an observed test result.
What to inspect: A governed analytics surface with explicit role boundaries. No Snowflake account or query was accessed here.
Workflow reference ↗Read these constraints before granting access.
A server supports a maximum of 50 tools; Snowflake notes that higher tool counts can degrade tool-selection accuracy.
Boundary reference ↗The managed server does not support several MCP constructs, including resources, prompts, roots, notifications, negotiation and sampling.
Boundary reference ↗Make the expected output explicit before you connect.
A governed analytics surface with explicit role boundaries. No Snowflake account or query was accessed here.
Keep the task inputs, tool calls, output and any errors together when you evaluate. The trial above is a proposal, not a recorded result.
Specific reasons to choose another route, with the tradeoff shown.
| Alternative | Choose it when | Tradeoff |
|---|---|---|
| Databricks MCP | Editorial choice: analytics and unstructured retrieval run in Databricks Unity Catalog. | Managed servers are currently public preview. Documentation ↗ |
| PostgreSQL MCP | Editorial choice: a local, read-only SQL inspection is enough. | Archived and lacks Snowflake governance. Documentation ↗ |
Answers specific to this implementation.
No. Snowflake documents a 50-tool maximum per MCP server and recommends additional servers when needed.
Documentation ↗
By default it uses the connecting user's DEFAULT_ROLE; Snowflake recommends restricting the MCP access role and secondary roles.
Documentation ↗
Maintainer documentation used in this record.