ClickHouse MCP ServerUnclaimed

ClickHouse MCP Server connects a ClickHouse database to AI assistants that support the Model Context Protocol. The assistant can list databases and tables, run SQL queries, and check a query's output schema or estimated reads before running it.

The server provides tools such as run_query, list_databases and list_tables. Queries run in read-only mode by default, and writes can be enabled explicitly with the CLICKHOUSE_ALLOW_WRITE_ACCESS setting. Query values can be passed separately from the SQL using ClickHouse's typed placeholders. A separate chDB tool runs SQL against files, URLs and other sources using chDB's embedded ClickHouse engine, once the optional chdb extra is installed.

It is for developers and teams who already use ClickHouse and want an AI assistant to answer questions about their data. The server works with the permissions of the configured ClickHouse user, so the assistant only sees what that user can access.

For HTTP and SSE transports, the server requires authentication by default, using either a static bearer token or OAuth or OIDC through FastMCP. A health check endpoint at /health reports whether the server can reach ClickHouse. The stdio transport needs no authentication and is the default.

The project is open source under the Apache-2.0 license and is hosted on GitHub. The page does not describe a hosted plan, so users run the server themselves. The README refers to uv and a Dockerfile for setup, and configuration is done through environment variables.