hypequery ClickHouse
Let agents query named ClickHouse datasets and metrics without handing them raw SQL
- Platforms
- CLI, API
- Runs
- Locally
- Version
- 0.15.0
- Listing updated
- today
Similar MCP servers
About hypequery ClickHouse
hypequery ClickHouse is an MCP server for databases. It runs locally on your machine and is started by your AI client.
The server draws a hard line around what an agent can touch. Only datasets you register and metrics you attach to them are exposed, and agents can inspect dimensions, labels, descriptions, filters, tenant keys and time keys before querying. The registry description sums it up: named datasets and metrics as tools, never raw SQL.
It runs over stdio by default, with an --http flag for a URL. npx hypequery mcp serves the datasets of a TypeScript API, hypequery mcp app:api does the same for a Python backend, and a --client-contract-url option points at an API that is already deployed. If a dataset has a tenantKey, the server refuses to start without a trusted tenant such as --tenant acme, and --self-test checks the setup first.
The docs are clear that MCP here is a capability boundary, not an authentication layer, so multi-tenant data needs the safety model read first.
Install
Run in a terminal
claude mcp add hypequery -- npx -y @hypequery/mcpAdd to claude_desktop_config.json
{
"mcpServers": {
"hypequery": {
"command": "npx",
"args": [
"-y",
"@hypequery/mcp"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"hypequery": {
"command": "npx",
"args": [
"-y",
"@hypequery/mcp"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"hypequery": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@hypequery/mcp"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"hypequery": {
"command": "npx",
"args": [
"-y",
"@hypequery/mcp"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"hypequery": {
"command": "npx",
"args": [
"-y",
"@hypequery/mcp"
]
}
}
}Using several servers? Build one config file with the MCP config generator.
What it does
- Governed datasets
- Exposes only registered datasets and attached metrics.
- Schema inspection
- Agents can read dimensions, labels, filters, tenant keys and time keys.
- Tenant guard
- Refuses to start without a trusted tenant when a dataset has a tenantKey.
- TypeScript and Python
- Serves datasets from either a TypeScript or a Python hypequery API.
- Self-test mode
- Validates the setup and counts exposed endpoints without running queries.
- Stdio or HTTP
- Defaults to stdio; add --http to serve at a URL.
Good for
- Ask Claude or Cursor for revenue by region using the same metric your dashboards use
- Give an analyst agent access to ClickHouse without a SQL console
- Expose multi-tenant analytics to an agent as one trusted tenant
- Print an mcpServers config entry with --print-config
hypequery ClickHouse FAQ
Can the agent run arbitrary SQL?
No. It works through named datasets and metrics, and the API keeps ownership of SQL.
Does it replace authentication?
No. The docs say MCP is a capability boundary, not an authentication layer.
How do I check my setup?
Run hypequery mcp with --self-test. It validates the contract and counts the endpoints it would expose without calling any query tool.
Is there a hosted option?
The overview lists hypequery Cloud, where every release has an /mcp endpoint.
How do I add hypequery ClickHouse to Claude Code?
Run this command in your terminal: claude mcp add hypequery -- npx -y @hypequery/mcp
What are the best alternatives to hypequery ClickHouse?
The closest alternatives to hypequery ClickHouse are Agent-Native Analytics, Google Ads + Meta Ads + SEO/GEO MCP, dashboard-icons, Tapetide - Indian Stock Market MCP and Midplane.
Where this comes from
- Description: written from the project's README and the project's website.
- Install steps: from the official MCP Registry entry.

