Snowflake

This is Snowflake's official MCP server for running SQL queries against your data warehouse from Claude. You authenticate with account credentials, and Claude can then execute analytical queries, inspect schemas, and pull results directly into the conversation. It's built for data teams who want to let Claude act as a conversational query layer over their Snowflake instance. I'd reach for this when I'm doing exploratory analysis and don't want to context-switch between Claude Desktop and the Snowflake web console. Ask Claude to "show me revenue by region for Q4" and it writes the SQL, runs it, and formats the results. The workflow shines when you're iterating on queries or need to explain data patterns to non-technical stakeholders. Claude can refine queries based on your feedback without you touching the SQL. The setup is straightforward if you already have Snowflake credentials. You drop three environment variables into your Claude Desktop config and you're running queries. The server handles connection pooling and query execution. It doesn't try to be clever about caching or query optimisation, it just runs what Claude asks for. Quirks: this is read-only by design, which is sensible but means you can't use it for data transformations or table creation. Claude also can't see query plans or performance metrics, so it's not useful for optimisation work. The server assumes you trust Claude with your warehouse credentials, which is fine for personal projects but needs careful thought in a team setting. There's no row limit enforcement in the server itself, so a poorly scoped query could pull back a massive result set. Skip this if you're not already a Snowflake customer or if your use case is just occasional lookups. The value is in the conversational iteration, not one-off queries. Also skip if you need write access or want to use this in production pipelines. This is an analysis tool, not an orchestration layer.