WWDC26: the year agentic apps moved on device
What actually shipped?
When people say “the agentic era,” they tend to picture cloud systems booking flights and answering email unattended. WWDC26 pointed somewhere quieter and, for personal work, more important: agents that run on the device.
Two announcements did the work. First, the Foundation Models framework — Apple’s on-device models, opened to every developer, callable locally at no cost with nothing sent to a server. Second, a Siri rebuilt around App Intents, so any app can expose its actions to be invoked by name through Siri, Shortcuts, and the Share Sheet.
Read together, they are a starter kit for agentic apps. One gives an app intelligence that stays on the phone. The other gives you a way to call that app the way you would call a person: “…with my invoicing app,” “…in my writing agent.”
Why is this the on-device turn?
Because the whole loop can now close locally. Before, an app that wanted real generation had to reach a cloud model, which meant an account, a server, and your data leaving the phone. After WWDC26, a focused app can do its job — draft, read, rewrite — entirely on the device, then use App Intents to be summoned in the moment.
That reframes what an “agent” has to be. It does not have to be a background worker roaming your accounts. It can be a small, private specialist that lives on your phone, holds your rules, and acts only when you ask. For invoices, receipts, and messages, that is the version worth wanting.
Where does Ilura fit?
Squarely in this lane, and before the keynote made it fashionable. Ilura Technology builds on-device agentic apps for companies of one — ILURA for the words you write, Keel for the money you run. Both run their generation through Apple Intelligence, both are invoked through Siri and the Share Sheet, and both keep every input on the phone. WWDC26 did not change the plan. It handed the whole approach a much bigger stage.
Read the signal through ILURA
Platform news matters when it changes what users expect from personal AI. ILURA reads these shifts through one lens: private agents trained by the user on iPhone.
- What becomes possible?
- What should stay user-controlled?
- What belongs on device?
Try it now
Put this to work on a real message.
Open ILURA, bring in a message you actually need to handle today, and get it done in your voice — free, on device, no account. It learns the preference, so the behavior carries to the next one.
Free to start · No account · Data Not CollectedQuick answers
- What at WWDC26 makes agentic apps possible?
- Two things. The Foundation Models framework lets any app run Apple's on-device models locally, for free, with no server. App Intents lets Siri, Shortcuts, and the Share Sheet invoke an app's actions by name. One provides the intelligence, the other the way to call it — that is the shape of an agent.
- Are these agents autonomous?
- Not in the runaway sense. The pattern Apple encourages is user-invoked and reviewable: you ask, the app acts, you confirm. That is a deliberately narrower kind of agent than a cloud system acting unattended, and for personal data it is the safer one.
- Why does on-device matter for agentic apps?
- Because an agent that does your invoicing or your messages handles data you would not hand to a stranger's server. Running on device keeps that data on the phone, works offline, and needs no account — the difference between an agent you own and one you rent.