Feature Flag

You might say

Open the new recommendation feature to employees first; if error rate rises, turn off the new path immediately without redeploying.

A feature flag enables or disables a code path at runtime according to configuration or user conditions.The same deployed code can send employees to a new recommendation page while other people stay on the old path; closing the flag stops the new path on anomaly. It separates deploying code from deciding who gets a feature. Permanently deleting code or restoring a whole old version is not a feature flag itself.

Know first

Deployment State

Feature FlagFeature ToggleFeature Gate

How the same code shows a new feature to employees and the old path to others

feature / smart-itineraryNew itinerary recommendations

Employee · lin@trip.local

Classic itinerary listBrowse by date

Customer · guest@example.com

Classic itinerary listBrowse by date

Deployment and exposure are separate; the rule decides who enters the new business path.

What a flag separates

Deployment and exposure can happen at different times: Code can already be deployed, while a flag sends only employees, a small group, or qualifying people to the new path. Observe real signals, then expand without waiting for another code release to control access.

Turning a flag off quickly stops one path: When new recommendations misbehave, turning off the flag returns people to the old path; monitoring, root-cause repair, and verification still follow. A flag does not repair data or restore a whole version.

You can say this to an AI Agent

Add a feature flag for the new recommendation page: enable it for employees only by default and keep other users on the old page. Monitor the new-path error rate after release; above threshold, disable the flag and confirm everyone returns to the old path without rolling back the entire version.

Learn next

1Monitoring→2Rollback→

Recommended tool

Web fine-tuning tool · Selector Pro

Select page elements for precise AI edits

Click a button, text, or card, note the change, then copy its exact location to AI.

Open Selector Pro website

The new recommendation page is deployed, but the team wants only employees to try it first and everyone to return to the old page immediately on error without redeploying. What fits best?

A. Add a feature flag for the new path, enable it only for employees, and disable it on anomalyB. Roll back the entire version whenever the trial audience needs adjustmentC. Permanently delete old-page code first, then restore it if there is a problem