Get the list of flags still on in production every Monday

Audit every GrowthBook flag as a table

Fetching a single feature returns the whole object: every environment override and the full rule chain, not just an enabled state. Run that across the project and the flag audit is a table rather than half an hour of clicking.

Listing pages on limit and offset with a default of a hundred, so a large project pages rather than truncating. Put the feature list beside the experiment list, which carries id, name, hypothesis, status, variations and metrics, and the flag audit and the test inventory fall out of the same pass. That is data extraction with a release-notes flavour.

What does the toggle actually change?

One thing: whether a feature is enabled in one environment.

It leaves rules, rollout percentages, default values, targeting conditions and every other environment untouched. A feature on behind a ten-percent rollout goes fully off, and comes back on with that same rule chain intact.

The tool calls production toggles high impact and asks you to confirm before flipping them, and the product enforces the same thing: a consequential call waits for your explicit approval. The environment id comes from the environments list, so “production” is a value the companion read from your account rather than a string it assumed.

What can a companion reach in GrowthBook?

Every feature with its rule chain, the environments, the projects and the experiment inventory, plus the single toggle. Weekly, that covers most of what a flag estate needs: the audit, the launch checklist naming what remains gated, the pre-release diff between staging and production.

Authoring lives in GrowthBook: creating features, editing rules, moving rollout percentages, attribute schemas, SDK connections, saved groups, namespaces and webhooks. The experiment list returns the design rather than the numbers, so results are read in GrowthBook where the statistics engine sits.

• Features: list with paging, get one with every environment override and rule.
• Environments and projects: list, so ids come from the API rather than from memory.
• Experiments: list metadata (hypothesis, status, variations, metrics). No results.
• The one write: toggle a single feature on or off in one environment, after approval.

Connect GrowthBook with a Personal Access Token

A Personal Access Token carries the permissions of whoever minted it, so mint it as an account whose access matches the job. Read-only is a perfectly good answer here, and the five read tools carry most of the value.

The integration talks to GrowthBook’s hosted API, so a hosted account connects and a self-managed instance stays on its own host. Check which one you run before you plan a routine around it.

Frequently asked questions