Our companies
What is a feature flag?
A feature flag, also called a feature toggle, is a switch in your code that turns a feature on or off, for everyone or for chosen users, without a deploy. You ship the code switched off, turn it on for a few people, widen it, and turn it off in seconds if something breaks.
Updated November 8, 2026
How does a feature flag work?
The code asks a question before running the new path: is this flag on for this user? The answer comes from a flag service or a table you control, not from the code, so changing it doesn't need a release. A good flag service evaluates flags inside your app from a cached ruleset, so a check is fast and doesn't depend on the network. In Flagger.gg, a flag that is unknown, not loaded or errored counts as off, so mistakes fail safe.
What kinds of feature flags are there?
The code looks the same for all of them; what differs is how long the flag lives and who changes it. Flagger.gg's explainer sorts them like this:
| Kind | Example | How long it lives |
|---|---|---|
| Release flag | A rebuilt checkout ships dark, then goes to admins, 10% of users, then everyone | Days to weeks, then delete it |
| Kill switch | PDF export turned off at once when a bug hits, while the fix ships | As long as the feature can fail |
| Beta or permission flag | Bulk invoices on for three named customers only | Until the feature is for everyone |
| Agent change flag | An AI coding agent merges a change behind its own flag, and a person turns it live | Days, then delete it |
| Experiment flag | Two variants compared on a metric | Experiments are in development in Flagger.gg |
Who can a flag be on for?
Flagger.gg has five states: off, admins only, a list of targets (named users and workspaces, plus admins), a sticky percentage of users, and everyone. A percentage rollout keeps each user's answer stable as you widen it, so nobody flips back and forth.
What does feature flag management mean?
A single flag is an if statement. Management is everything around many flags used by many people and, more and more, by AI agents:
- one place to change state per environment;
- targeting by user, workspace or percentage;
- permissions, such as Flagger.gg's agent tokens scoped to one project, chosen environments and actions;
- an audit log of who changed what, when and why;
- flags declared in code, so the code and the service agree; see keeping AI-written code behind declared flags;
- cleanup once a feature is fully on.
Why do AI-native teams lean on flags?
Because agents ship many small changes quickly. A flag around each change lets it merge and deploy switched off, and lets a person or the agent turn it off in seconds without a redeploy. North Bay Ventures ships this way instead of using a staging server; read feature flags instead of staging and trunk-based development with AI agents. Flagger.gg is built from the flag system that runs Sandboxer and Transformer in production.
Frequently asked questions
Is a feature flag the same as a feature toggle?
Yes. They are two names for the same thing: a switch in code that turns a feature on or off without a deploy.
Do feature flags slow an app down?
Not when the flags are evaluated in the app from a cached ruleset, as Flagger.gg's Go and Node SDKs do. The check doesn't call the network.
When should you delete a feature flag?
A release flag should go once its feature is on for everyone and stable. Kill switches for features that can fail are meant to stay.
Can AI agents change feature flags?
In Flagger.gg, yes, through MCP, the CLI or the API with a scoped token. Every agent write needs a note, and agents have no approve tool.