apple is putting ai agents on the app store. the interesting part isn't the agents — it's what apple has to do to its own rulebook to let them in.
what happened#
a report this week (via 9to5mac, may 13) says apple is working on a way to formally incorporate ai agents into the app store — software that takes multi-step actions on a user's behalf. the likely venue for an announcement is wwdc next month, though apple may not be ready to ship it.
the reason this is hard for apple is the reason it's interesting for the rest of us. agents break the assumption app review is built on. review checks a binary: does this build do what it says, does it stay inside the guidelines, does it avoid the things apple bans. an agent's behaviour isn't fixed at submission time. it's decided at runtime, by a model, in response to a prompt apple never saw. apple can't review what it can't predict.
why it matters for indie devs#
if you ship an app with an agent in it — and in 2026 a lot of you do — you're about to be on the wrong side of a guideline that doesn't exist yet. apple will write rules for agentic apps. it always does. and the first version of those rules will be blunt, because the first version always is. think of the 5.1.2(i) ai-consent rule, or the 2.5.2 crackdown on apps that finish building themselves: both landed fast, both rejected apps that were acting in good faith, both got refined only later.
the apps that survive that first wave won't be the ones with the cleverest agent. they'll be the ones whose submission is legible — whose metadata, permissions, and declared behaviour line up with what the app actually does, so a reviewer (or apple's own review agent — the report hints review itself gets automated) can map the app without guessing.
the stora angle#
this is the boring half of shipping an agentic app, and it's the half that gets you rejected. your app's capabilities drift from its store listing every release. its privacy declarations lag the sdk you actually pulled in. its description promises an agent that does X while the binary does X and Y.
that's the gap stora's compliance engine is built to close — it reads your actual build and flags where the listing, the permissions, and the declared behaviour disagree, before apple's reviewer does. when the rules for agents land, the work isn't rewriting your agent. it's proving your submission says what your app does. that's the thing stora keeps in sync.
the bigger point#
apple isn't slow on agents because the tech is hard. it's slow because its entire review model assumes an app's behaviour is knowable at submission time, and agents quietly broke that assumption. whatever apple announces at wwdc, the cost lands on developers as a new compliance surface — not a new feature.
so don't wait for the rules. go read your store listing against your actual build today. the gap you find is the gap apple's about to start rejecting for.