The three frameworks, in one paragraph
FamilyControls asks the user for permission and shows Apple’s app picker. ManagedSettings applies the restrictions: the Shield screen you see when you open a blocked app. DeviceActivity runs schedules and wakes a small extension of your app when an interval starts or ends, or when usage passes a threshold. Your main app is usually not running when any of this matters.
1. Your app is asleep when its own blocks start
A schedule starting at 22:00 doesn’t launch your app. iOS calls your DeviceActivityMonitor extension, and that extension has to apply the right shields by itself, from whatever state you saved in a shared app group.
Two things surprised me:
- Callbacks arrive late. A start is usually delivered a few seconds after its minute, and later if the phone hasn’t been used. Code that asks “which block is running right now?” has to accept a start that shows up after its minute, and for a night block, after midnight.
- Callbacks can arrive at the same time. When one schedule ends at the exact minute another starts, iOS can call
intervalDidEndandintervalDidStartat once on two threads. Both read, change and save the same blocking state. The open-source blocker Foqos tracked a version of this in issue #603. My fix was boring: every callback in the extension runs under one process-wide lock.
2. iOS will lose your monitors, so build a repair pass
Monitors disappear. Turning Screen Time off and on is one way. Once a monitor is gone, nothing tells you, and the block simply doesn’t start, or worse, doesn’t end.
So EvoCat runs a repair pass every time it comes to the foreground. It compares three things: what the user saved, what the blocking state says, and what DeviceActivityCenter reports as registered. Then it registers what’s missing.
The rule that took me longest to learn: never stop a monitor that is still registered just to “refresh” it. Stopping a running monitor ends its interval. That can lift the Shield in the middle of a block, or end a break early. The repair pass only adds what’s missing and leaves everything that’s still there alone.
The shields have a second safety net. Whenever the app or the extension re-applies them, it checks each saved window against the clock and corrects any day still marked “on” whose end callback never came.
3. You get twenty schedules, and the number is easy to miss
startMonitoring throws excessiveActivities when you register too many. The number isn’t on the DeviceActivityCenter page; it’s on the error’s own page: twenty activities at a time, for your app and its extensions together.
My first design registered one monitor per selected weekday per block. Two every-day blocks used 14 of the 20, leaving no room for a third block plus a daily-limit monitor and a break timer. Since November 2026 each repeating block uses one daily activity, and the extension decides whether today is one of the days the user picked.
If you’re starting fresh, treat monitors as a scarce budget from day one, and keep a little headroom for short-lived ones like breaks.
4. Intervals have a fifteen-minute minimum
The same goes for intervalTooShort: its page says the minimum interval is fifteen minutes. EvoCat pads anything shorter by moving its start earlier.
The padding has its own trap. A ten-minute block at 00:05, padded backward, starts the evening before, on a day the user never selected. The app now rejects a same-day window that padding would push across midnight.
For a short timed opening, like letting someone use an app for five minutes, the trick is a one-off schedule whose start already lies in the past: iOS starts it at once and ends it when you asked. One more detail: the schedule counts whole seconds, so round the end up. Rounded down, it could end up to a second before your app thought the opening was over.
5. Night blocks and time zones
DeviceActivitySchedule uses hour, minute and weekday with no time zone. After a flight from Berlin to New York, iOS keeps running a 22:00–07:00 block at 22:00–07:00 local time. That’s what people expect.
My app, though, had saved those times as absolute dates. Read in New York, a 09:00–17:00 Berlin block looked like 03:00–11:00, and the “end any block that should be over” safety net from section 2 would have ended a running block hours early.
The fix moves every saved time to the same wall-clock time in the new zone. Because the app can be killed halfway through, it writes the whole move down first, applies it store by store, and only then remembers the new zone. A move that was cut short is finished on the next launch, and nothing moves twice. Countdowns and breaks aren’t moved, because they’re lengths of time, not clock times. Until the move is done, the safety net leaves the clock-based corrections to iOS’s own callbacks.
7. You can show screen time, but your app can’t read it
To show usage you build a DeviceActivityReport extension. Apple’s documentation says it runs in a sandbox that prevents network requests and moving sensitive data out of the extension. In practice, anything it writes to your shared app group is silently dropped and never reaches your app, your widget or another extension.
So the report can show today’s screen time, and your app still can’t use that number for anything else. If you plan a feature that reacts to someone’s usage, design around this before you build it.
8. The simulator can’t tell you if blocking works
Screen Time behaviour only means anything on a real phone. EvoCat’s release test is a checklist run on a device by moving the clock by hand. It covers a night across midnight, a break that crosses midnight, a restart during a block, a time-zone change and a daylight-saving night.
One tip: set the clock two minutes before the moment and let it tick over. If you jump straight past a start or an end, iOS can miss it or deliver it late, and you’ll chase a bug that isn’t there.
9. What no blocker on iOS can promise
ManagedSettings can stop apps being deleted while a block runs; EvoCat turns that on only during an active hard block. That makes giving up harder, not impossible. Be suspicious of any app that calls itself bypass-proof.
Why I’m writing this
I build EvoCat on my own. It started in December 2024, and almost every rule above began as a bug report or a failed test. If you’re building on these frameworks and hit something I didn’t cover, I’d like to hear about it at [email protected].
Sources
- Apple: DeviceActivityCenter.MonitoringError.excessiveActivities ↗
- Apple: DeviceActivityCenter.MonitoringError.intervalTooShort ↗
- Apple: ShieldActionResponse.openParentalControlsApp (iOS 26.5) ↗
- Apple: DeviceActivityReport ↗
- Foqos issue #603: second scheduled profile doesn’t start when scheduled right as the first one ends ↗
Implementation details come from EvoCat’s own iOS source code. Send a correction