Chapter 02 · 4 min

Connecting GitHub and choosing repositories

The account is the GitHub sign-in. The permission model is the first thing an engineer checks.

There is no separate signup. Signing in with GitHub creates your organisation and your first workspace, then sends you to install the Trevo Bot GitHub App. The install page opens in the same tab — a full-page redirect, not a popup — and the workspace links itself when you come back.

Choose “Only select repositories”

It is pre-selected, and it is the recommendation. Trevo needs read access to the code it proposes against and write access to open branches named trevo/*. It never writes to main and never merges. Granting one repository is enough to get the first proposals; more can be added later from the same install page, and each repository becomes its own workspace with its own data.

What happens the moment a repo connects

The scan starts server-side and streams to the dashboard: files read, framework detected, funnel surfaces found. On a store like Terra & Twine that is 42 files and about a minute. If the repository already fires events through the SDK, Trevo reconciles them into a funnel. If it does not, Trevo opens one pull request that installs the SDK and wires the events your funnel needs. No experiment is generated until you ask for one.

Handing the step to an engineer

If you are not the person who can install a GitHub App, the connect screen has an “invite an engineer” action that sends an organisation-admin invite. They land on the same screen, install, and the workspace is ready for both of you.

What Trevo reads, and what it never touches

• Reads code from a temporary clone, then discards the clone.
• Reads and writes only on trevo/* branches.
• Never merges. Never deploys. Never ships a variant to 100%.
• Never trains a model, ours or anyone else’s, on your code.