Introduction

hypequery cloud will be the hosted home for your ClickHouse semantic layer. It is not live yet; this section describes how it will work. You define datasets once: the dimensions people can group by, the measures they can aggregate, the relationships between tables, and who is allowed to query them. Cloud then serves those datasets to your dashboards, applications, and AI agents, without a server for you to run.

Until cloud launches, serve the same datasets yourself with the open-source SDKs: see the TypeScript and Python quick starts.

Author datasets your way

Cloud accepts datasets from two authoring paths. Both produce the same deployment contract, so cloud serves them the same way.

• With AI. Cloud's AI onboards you, learns your business, and drafts your datasets for you to review. See AI authoring.
• In code, with TypeScript or Python. Define datasets with the open-source SDKs, review them in pull requests, and deploy with the hypequery CLI or from CI. See Deploy from TypeScript and Deploy from Python.

Code-defined datasets are not tied to cloud. The same definitions run inside your own app with the open-source SDKs, so you can start self-hosted and move to cloud later, or the other way around.

What cloud serves

Every release is served two ways:

• REST endpoints, one per published dataset, for dashboards, apps, and integrations. See REST API.
• An MCP endpoint, so AI agents can discover and query the datasets you publish. See MCP.

Access is set when you publish: you choose the roles and scopes callers need for each dataset, and cloud checks them on every request. For customer-facing analytics, API keys also decide which customer tenant each request runs as, and releases can be rolled back in one click.

Next steps

• Quick start: from sign-up to a live endpoint
• How it works: from dataset definitions to a live release
• Deploy from TypeScript
• Deploy from Python