Database MCP servers
These servers let an AI assistant query and manage real databases: run SQL, inspect schemas, and answer questions from live data instead of guesses. Most support read-only modes, which is the right default for anything an agent touches.
What engineers use them for
- Ask questions in English and get answers computed from production data
- Generate and validate SQL against the actual schema before it runs
- Let an agent inspect tables and indexes while debugging a slow query
- Automate routine reporting straight from the source of truth
Highest-graded right now
Graded on maintenance, completeness, and reliability. The formula is public →Supabase
MCP server for interacting with the Supabase platform
ArcadeDB MCP Server
Built-in MCP server for ArcadeDB multi-model database (graph, document, vector, time-series)
QueryWeaver
An MCP server for Text2SQL: transforms natural language into SQL using graph schema understanding.
ClickHouse
Official ClickHouse MCP server for querying and exploring ClickHouse clusters and chDB.
Postgres
Enterprise PostgreSQL MCP server with NL queries, hybrid search (pgvector+BM25), and web UI
mcp
Connect your AI assistants to Keboola and expose your data, transformations, SQL queries, ...
Dynoxide
DynamoDB-compatible database engine in Rust on SQLite, with an MCP server for coding agents.