Quickstart
1. Link your Semantic Repository
Under Repository, link a GitHub repository. It can be empty: qontexta writes
the Cube project layout into it (model/cubes, model/views, cube.js, a README)
without touching anything that was already there.
Two ways to grant access:
- GitHub App (recommended) — install the app on your account and paste the installation id. Tokens are minted on the fly and expire within the hour: no long-lived secret is stored here.
- Personal access token — needs write access to contents (
contents:write). Stored encrypted and never shown again.
2. Connect your Data Platform
Under Connections, add your database. Read-only permissions are enough: qontexta reads schemas and runs queries, never writes.
3. Tell it what you want to measure
Under Assistant, pick the connection, narrow it to a schema and say it in your own words: “I want to measure revenue by region and average order value”.
The assistant reads your tables and proposes cubes, measures and semantic views, with the reasoning for each and the questions it could not decide on its own. You review, ask for changes if needed, and what you accept is written in a single commit.
You do not bring an AI key: the assistant is included and runs on qontexta’s own account. Your plan comes with a monthly allowance.
4. Sign the meaning
Under Model, open a metric. You will see two tabs: Business and Technical. Fill in what it measures, what it leaves out and who answers for it, then validate. Without a title, a description and an owner it cannot be validated — validating something unexplained would claim a number nobody has described is trustworthy.
5. Connect an agent
Under Environments, hit Connect on production. There you will find the MCP endpoint, the semantic SQL one, REST and GraphQL, with a token scoped to you.