Shipeasy
Flags & ExperimentsCase studies

QA a flag before you ramp (and share a repro link)

Use the devtools overlay (Shift+Alt+S / ?se=1) and ?se_gate_… URL overrides to QA a feature, demo a variant, and reproduce a bug — all session-scoped.

Production readyOn this page · 3 min readUpdated · June 19, 2026Works with · @shipeasy/sdk · client

A flag is built but sitting at 0%, and QA needs to verify it before you ramp. You don't want to flip the dashboard (that's everyone) and you don't want a special build (that's a deploy). You want to see the feature on your own screen, right now — and ideally hand a teammate a link that reproduces the exact state.

The browser SDK ships a devtools overlay for precisely this: flip any gate, config, experiment, or translation for your session only.

Open the overlay

  • Press Shift+Alt+S on any page running the SDK, or
  • append ?se=1 to the URL.

It mounts in a Shadow DOM overlay, lists every resource in your project's blob, and lets you toggle each one. Your changes apply to your session and nobody else's.

Session-scoped, client-side only

Overrides live in your browser for the session — they never write to the dashboard and never affect other users. Close the session and you're back to the real evaluation.

For QA flows, demos, and bug reports, force state straight from the URL — paste the link and the receiver lands in the same state:

ParamEffect
?se_gate_<name>=trueForce a gate on/off
?se_config_<key>=<value>Force a config value
?se_exp_<name>=<group>Force an experiment group

For example, https://app.example.com/checkout?se_gate_new_checkout=true shows the QA reviewer the new checkout while it's still at 0% for real traffic.

These sit between programmatic overrides and the server's evaluation in precedence:

programmatic override  >  URL / devtools override  >  the server's evaluation

Rollout & measurement plan

  • QA at 0%, then ramp. Verify the feature on your own session before any real user is in the rollout — the URL override is your pre-flight check.
  • Demos without a deploy. Flip an experiment variant live in front of a customer, then close the session.
  • Bug repro links beat screenshots. A ?se_gate_… link reproduces the exact state for whoever opens it, so the bug is reproducible instead of described.
Was this page helpful?
✎ Edit this page

On this page