The MCP stack for data analysis
Warehouse, database and spreadsheets side by side, so questions get answered from the source.
Each role takes the best-ranked match: reviews first, then real adoption, then the safety grade, graded A or B, one per publisher. Nobody chose these by hand and none paid to be here. Recomputed 2026-11-03. How ranking works
1Warehouse
Large tables without exporting them.
Read-only BigQuery tools for plain-language data questions, with a cost gate on every query
No reviews yetlocal · pypi40/wkFree
2Warehouse, alternative
The other common warehouse.
MCP Server for Snowflake from Snowflake Labs
No reviews yetlocal · pypi29912k/wkFree
Also good: Snowflake MCP Server (B)
3Database
The operational data, read-only.
Comprehensive PostgreSQL documentation and best practices, including ecosystem tools
No reviews yetgateway-ready1.9k52/wkFree
Also good: Ekwo OS — open source accounting: ledger, invoices and VAT on your own Supabase or Postgres (A), PostgreSQL (B)
4Spreadsheets
Where the business already keeps numbers.
Google Sheets for agent workflows in Claude Code and Claude Desktop: APIs, files, schedules, undo.
No reviews yetlocal · npm35497/wkFree
Also good: Google Sheets (B), gogcli-mcp-sheets (A)
5Dashboards
The metrics the team already looks at.
MCP Server for Metabase - 111 tools for SQL generation, dashboards & PostgreSQL
No reviews yetlocal · npm4959/wkFree
Run it through one connection
Database is remote, so with the gateway added once your agent can call it by name. The others run on your machine: their pages have the install config.
claude mcp add --transport http mcp-market https://gw.mcp.market/mcp