Server detail
mcp-clickhouse is a Model Context Protocol server that connects ClickHouse to your MCP-compatible tools and agents. It lets you explore schemas, run SQL queries, and manage tables directly from an MCP client while enforcing safety controls: read/write/destructive classification, database allowlists, row caps, dry-run, and audit logging. Ideal for data analysts, ML engineers, and backend devs who want to query ClickHouse through AI assistants without exposing raw credentials or running risky statements. Integrates easily with any MCP host (Claude Desktop, custom agents) via standard MCP transport. Compared to ad-hoc DB clients, it adds structured governance for LLM-driven workflows—preventing accidental deletes, capping result sizes, and logging every action. Setup is straightforward: configure the server with your ClickHouse connection, define allowlists and limits, and start querying securely through your AI toolchain.
An MCP server that exposes mcp-clickhouse capabilities to MCP-compatible AI clients.
Collections featuring this MCP
Tool testing
mcp-clickhouse
Call the MCP capabilities provided by mcp-clickhouse and return a structured result.
inputPass arguments according to the server tool schema.Connection modes
{
"mcpServers": {
"mcp-clickhouse": {
"url": "Generated by the provider after deployment"
}
}
}The Remote endpoint is generated by the provider after deployment; this page does not fabricate an unusable endpoint.
No npm package is recorded. Open the source repository to complete command and args.If no npm package is registered, follow the installation method in the source repository.
How to use
- 01Step 1
Review server capabilities and permission scope.
- 02Step 2
Copy the install command or JSON configuration.
- 03Step 3
Run a small connection test in your client.
- 04Step 4
Adopt it long term only after reviewing access and maintenance.
Discussions
Use this space to keep checking source information, usage experience and maintenance status.
Open source page