Snowflake MCP Server
Snowflake Labs maintains this MCP server, wiring the warehouse into your AI: inspect table schemas, run SQL, and call Cortex's AI analytics. For colleagues who don't write SQL, it becomes a direct way to ask the data a question.
Compared with rolling your own query endpoint, the advantage is that the AI reads the schema before writing SQL — you don't paste table definitions into the prompt, it looks them up.
What it does for you
Connects your AI to Snowflake — run analytical queries over your data warehouse in plain language.
Capabilities
- Run analytical SQL
- Inspect warehouse data and schema
Once installed, try asking your AI
Say this directly in Claude / Cursor or any client with this server connected
- ›“Break down last quarter's revenue by region”
Install / Connect
# No npm package — install from source.
# See the repo's README for the exact command:
# https://github.com/Snowflake-Labs/mcpFollow the official README; env and args vary per server.
Why use it
Good for: data teams on Snowflake, and for letting business colleagues self-serve.
Before connecting: Snowflake bills per query, and AI-written SQL can scan whole tables. Use a dedicated role with a statement timeout and a resource monitor — don't connect as an admin.
Similar / alternative servers
io.github.ErikEJ/SqlServer.Rules
Analyze T-SQL scripts for design, naming, and performance issues and prevent bad
ai.matih/mcp
The grounded data layer for any LLM: governed SQL, metrics, lineage and catalog
mcp-neo4j
Neo4j Labs Model Context Protocol servers
postgres-mcp
Postgres MCP Pro provides configurable read/write access and performance analysi
Frequently Asked Questions
Could it run up a huge bill?
It's a real risk. A dedicated role plus a resource monitor and statement timeout is the standard guardrail here.
Does it support Cortex?
Yes, the official server includes Cortex capabilities — see the repo README for specifics.