Siri AI Opens to Every Developer: Your App, One Sentence Away
On Nov 3, 2026, Apple opened Siri AI to every developer: any app can now plug into Siri through App Intents, chain actions with skills, and be called on in plain language. The catch — users get it in a few months, App Store submissions open in spring 2027, and the EU is left out for now.

On November 3, 2026, Apple posted a short announcement on its developer website: starting with the iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3 betas, a new Developer Preview lets any developer integrate their app with Siri AI. The weight of the sentence is not in its length but in the word "any" — until now, only apps that fit the limited scenarios Apple had defined could get into Siri; from this day on, in principle any app can hand its content and actions to Siri, to be called on with a single spoken request. For developers who have spent years building iPhone apps, this is the first time Siri has genuinely become an open entry point.
But hold the excitement. This is still only a developer preview, not a public launch. Apple states it plainly in the announcement: the preview will reach customers "in a few months", and App Store submissions open next spring. In other words, on November 3, 2026, what developers can do is start adapting and testing — not ship Siri-enabled apps to users right away. Understanding this release depends on keeping the distance between "can connect" and "can use" clearly in view.
What was actually released
The mechanism is App Intents. Developers express their app's actions through App Intents, and Siri AI can then call on them. According to Apple, any app can now tap Siri AI's newest capabilities: personal context understanding, onscreen awareness, and natural language. Users no longer need to memorise command formats; they can ask in their own words, and Siri AI carries out whatever custom action the developer has built — Apple's examples are checking a delivery, booking a reservation, and placing an order, all small everyday things that can be said in one sentence.
The genuinely new piece is "skills". For complex requests, Apple is introducing skills, which let developers teach Siri AI how to combine several actions into a single experience. As developer-community readings of Apple's documentation explain, skills are simple text files (SKILL.md) in which a developer describes in words how several actions chain together to complete a task. Apple is already shipping three skills of its own in the first iOS 27.3 beta, including one that teaches Siri how to select and edit text in the Notes app. That is a signal worth noting: Apple is not only opening the interface, it is using its own apps to demonstrate what "combining actions" should look like.
There is one more change that is easy to miss: Siri AI can now surface any entity in the Spotlight semantic index, not just ones that fit an existing schema. In other words, what Siri can talk about is no longer limited to the data types Apple predefined — an app's own content can enter Siri's understanding too. If this delivers, it marks the leap from Siri as a tool that "can operate a few fixed scenarios" to Siri as a personal assistant that "understands the specific things inside your app".
Apple's own 15 apps went first
Before opening up to third parties, Apple rolled these capabilities out across 15 of its own apps: Calendar, Files, Finder, Journal, Mail, Maps, Messages, Music, Notes, Photos, Preview, Reminders, Safari, Shortcuts and Weather. The announcement gives two examples: asking Siri to find a coffee shop you visited last week, or to archive files that have not been opened in over a year. Both point to the same capability: Siri knows not only what you just said, but what you have done in the past, and can act on that knowledge.
Language coverage is expanding at the same time. Siri AI currently supports English, French, Japanese, Korean, Portuguese and Spanish, and Apple says Traditional Chinese, Norwegian, Turkish and Vietnamese are coming soon. Looking back at the timeline: Siri AI was unveiled at WWDC on June 8, 2026, began rolling out as an English beta on September 14, and opened to all developers on November 3. Three steps in five months: first the reveal, then a small-scale user beta, then full developer access. The pace itself shows that Apple treats Siri AI as a system to be refined gradually, not a product fixed at launch.
The timeline: what you can and cannot do now
For developers, the timeline is this: you can now develop and test Siri AI integrations on the 27.3 betas across four platforms; the customer release comes "in a few months", with no specific date from Apple; App Store submissions open in spring 2027. According to the developer documentation, integrations built during the preview cannot be tested through external TestFlight, nor shipped in App Store releases — only built on a development machine with a development profile, for up to 100 internal testers via internal TestFlight. That means Siri integrations written this winter are essentially pathfinding prototypes, not features that can generate revenue right away.
There is also a list of constraints, all in the announcement's footnotes. First, Siri AI is currently unavailable in the EU on iOS, iPadOS and watchOS, and Apple says it is working constructively with the EU to find a path forward — developers in Europe cannot test this yet. Second, Siri AI is not available for users under 13. Third, it requires an Apple Intelligence-enabled device set to a supported language. Fourth, Apple Intelligence features that rely on server-side models, including Siri AI, are subject to daily usage limits that vary by feature, request complexity, system demand and other factors, and Apple says expanded access will be available for a fee in the future. As for the terms and usage policies for experiences enabled through Siri AI, Apple says "more to share soon" — developers still do not know the long-term pricing rules for this capability.
Why this step matters
To judge the significance of this release, look at what it changes. The old Siri integration ran on the SiriKit framework, with limited scenarios and developers boxed into the categories Apple had drawn. The new Siri AI integration runs on App Intents: any app can hand over its actions, and skills can chain several actions into one task. This is a structural change. The assistant stops being merely a voice shell for the operating system's own features and starts becoming a single entry point for every app. Users may no longer open an app and tap buttons first; instead they may speak to Siri first, and let Siri decide which app to call and in what order.
For Apple, this is also about closing a gap. Over the past few years, the competition in AI assistants has focused on the models themselves: whose model is smarter, whose agent can run longer tasks on its own. Apple's approach is different. It is not competing on model size, but on the entry point — the iPhone, Mac and Apple Watch are in users' hands every day, and if Siri can call on all the apps on them, its practical value may rival that of a smarter chatbot. The November 3 preview is the first step in handing control of that entry point to third-party developers.
But the bigger the entry point, the bigger the risk. On the same day, Apple's Form 10-K annual report filed with the US Securities and Exchange Commission added explicit AI agent risk language for the first time: agents may act in ways nobody directed. Apple CEO John Ternus reportedly emphasised around the same time that Siri AI does not run long autonomous sessions the way competitors do. Read together, the two documents make Apple's posture clear: it wants to open up, but within rules. One sentence in the developer documentation is worth remembering for everyone who integrates: Siri can only perform actions you make explicitly available. Developers draw the boundary of what is open themselves — and if they draw it wrong, the responsibility is theirs too.
China: a different track in the same week
Almost the same week, Apple took a different track in China. Apple's China newsroom announced on October 31 that the first Apple Intelligence features would enter developer testing in China from November 4, reaching the public with iOS 27.3 in December. The first batch includes Writing Tools, Live Translation, Visual Intelligence, Clean Up in Photos and notification summaries, with Alibaba's Qwen (Tongyi Qianwen) integrated into parts of the experience such as Siri, Writing Tools and Visual Intelligence. But the China package explicitly excludes the new Siri AI, as well as Image Playground, and the first release covers iPhone only — iPad and Mac await further approvals. In other words, what Chinese users get in December is the first generation of Apple Intelligence features, not the new Siri that opened to developers worldwide on November 3. The two tracks differ in both timetable and content, and developers planning for the Chinese market must count them separately.
This split also shows Apple's position in the Chinese market: whether a feature can ship depends not only on whether the technology is ready, but on how far regulatory approvals have gone. Apple has given no timetable at all for Siri AI in China. For developers in China, the practical move now is to integrate the first-generation Apple Intelligence capabilities — Writing Tools, Visual Intelligence and the rest — while watching for progress on Siri AI approvals, rather than assuming the global timetable will copy over unchanged.
Seen against Android
Apple is not the only company turning its assistant into an app entry point. Google's AppFunctions on Android does the same thing — letting the assistant call app functions — and is also currently in an experimental preview. The thinking is similar on both sides: neither is satisfied with an assistant that only chats; both want the assistant to act inside apps on the user's behalf. The difference is in pace and control: Apple opened up to every developer at once, but gated App Store submissions until next spring, leaving months to refine; Google is moving forward in a smaller experimental form first. Both approaches have trade-offs. Apple's lets developers start early, but the first integrations face a long wait, and enthusiasm may cool while waiting.
For developers, there is also a more practical comparison: how much of the experience built for Siri still applies? The answer is that the mechanism has changed, but the thinking has not. In the SiriKit era, developers had to fit their app's actions into Apple's preset scenario templates; in the App Intents era, developers define their actions clearly, then use skills to teach Siri how to combine them. Only the form of expression has changed. What has not changed is the core requirement: actions must be defined clearly enough for Siri to dare to execute them on the user's behalf. Features with fuzzy boundaries, where one wrong step is hard to undo — direct payments, deleting data — will most likely be left out of the first integrations by developers themselves.
One more detail is worth noting: Apple says users will "in many circumstances" not need to mention the app's name explicitly for Siri to call on it. That sounds convenient, but behind it lies a new dimension of competition: when users stop tapping app icons and instead speak to Siri, which app Siri picks becomes a new traffic entrance. App Store rankings used to decide exposure; Siri's choices may decide it too. Developers adapting to Siri AI now are not just adding a feature entrance — they are staking an early claim in the coming contest to "be picked".
A practical checklist for developers
If you want to start now, here is the order to follow. First, update to the iOS 27.3, iPadOS 27.3, macOS 27.3 or visionOS 27.3 beta, and confirm your device supports Apple Intelligence and is set to a supported language. Second, read Apple's new Siri AI developer documentation, focusing on the new App Intents APIs, how skills are written, and how to evaluate your app's experience with Siri AI — Apple specifically mentions evaluation methods, which signals that it wants developers not only to connect, but to measure accurately. Third, start with one or two of the most-used, least error-prone actions, such as queries and display actions, and push anything involving payments, deletion or sending things out to later, after the first round of feedback. Fourth, attend the Apple Intelligence Group Labs on November 4–5 and put your hands-on questions directly to Apple's people.
Two more things not to overlook. One is the daily usage limits: features that rely on server-side models have daily caps, and expanded use will cost money in future — factor that cost into product planning instead of assuming unlimited calls. The other is the uncertainty in the terms: Apple says it will follow up with details on the terms and usage policies for experiences enabled through Siri AI, and until that document appears, do not promise anyone a launch date or pricing model for a Siri integration. Code written during the preview may also change — Apple's documentation explicitly warns that the App Intents API and Siri functionality may change during the Developer Preview. What you write now is a probe, not an asset.
What "personal context" actually means
The sentence in the announcement most easily skimmed over — and actually the most important — is the first of Siri AI's new capabilities: "personal context understanding". According to Apple, Siri can understand not only what you just said, but what you want, drawing on your past emails, calendar, files and photos. The announcement's example of "find a coffee shop you visited last week" is personal context in action: Siri has to know where you went last week to answer. And "archive files that haven't been opened in over a year" requires Siri to know each file's last-opened time, then act on your instruction.
That capability raises a question that cannot be avoided: the more Siri knows, the higher the cost of a mistake. If it misunderstands context and archives a file it should not have, or sends private information to the wrong person, the consequences are far worse than a wrong answer. Apple's answer is built into the product design: Siri can only execute actions developers explicitly open up, and much of the processing is said to happen on-device. But the announcement itself does not spell out how personal-context data is stored, how long it is kept, or how users control it — those details await Apple's follow-up terms. For developers, that means treating "what if it misunderstands" as the first design priority: asking the user to confirm once before a critical action is far cheaper than apologising afterwards.
For ordinary users, nothing changed on November 3: your Siri is still the same Siri. But for developers, the rules of the game have changed. From now on, an app's value depends not only on how good its own interface is, but on whether Siri can call on it with a single sentence. Apple spent five months carrying Siri AI from the keynote stage into developers' hands; the next few months belong to developers, who must now prove whether this road leads anywhere.
Comments (0)
Sources
Related articles
On 17 November 2026, Microsoft launched MAI Code 2 and Code 2 Flash: the larger model has 77 billion active parameters and a one-million-token context, averaging 63.2% across four benchmarks — close to GPT-6.1 Sol and Claude Opus 5.5 — at $0.309 per task. Read the three caveats first: promotional pricing, the lowest score of the three, and no long-horizon work.
On 11 November 2026, AWS open-sourced UAEF: adapters turn traces from LangGraph, LangChain, Strands and Bedrock AgentCore into one format, then 36 metrics across 7 dimensions score them, flagging regressions past 5%. The judge model is hardened too — scores must cite verbatim evidence, and never decide safety.
On 17 November 2026, Perplexity launched the Orchestrator API: one interface reaching OpenAI, Anthropic, Google, xAI and open-weight models such as GLM, Kimi and DeepSeek, billed at the answering model's published rate with no platform fee. When one provider fails, the same model retries on another deployment, and preview routing assigns requests to models by task.