The Siri developer preview is now open to everyone

Apple opened the Siri AI Developer Preview to all developers on November 3, 2026. The preview is available through the iOS 27.3, iPadOS 27.3, macOS 27.3, and visionOS 27.3 beta releases. Apps can expose their own content and actions to Siri through App Intents.

Apple’s examples focus on everyday tasks: tracking packages, making reservations, and placing orders. Previously, Siri could only work with data that fit schemas Apple had defined within the system. The new preview removes that limit; developers now decide which content and which actions will be exposed to Siri.

Users don’t have to say the app’s name

Integrations can take advantage of personal context awareness and content visible on the screen. That lets users make requests in their own words. Siri can also surface every item in Spotlight’s semantic index, not just the ones that fit a predefined schema.

According to Apple, saying the app’s name is not necessary in every case. If the app is in the foreground, a Live Activity is open, or the Siri conversation already includes that app, saying the name is unnecessary. For background requests, though, the app name is required. In Apple’s example, a passenger following a Live Activity can ask Siri to change the pickup point without repeating the app’s name. Siri can also find content in one app and transfer it to a second app to take action there.

Developers choose which data to share

Apple wants access to be selective. A new or existing custom intent becomes a tool Siri can call through the @AppIntent(.tool) Swift macro. For intents in the primary domain that already match a schema Siri supports, this annotation is not required.

Titles and descriptions for intents, parameters, and entity properties help Siri interpret natural-language requests, distinguish between similar actions, and build a response. Developers can index entities in Spotlight, map on-screen content to entities, and donate intents when users perform similar actions to provide behavioral hints.

Access can be narrowed for sensitive data. Authentication policies can protect an entire entity or selected properties. In Apple’s finance example, authentication is required for the account nickname and balance, while the bank routing number stays inside the app and its extensions, and Siri never sees that field at all.

Skills for multi-step tasks

Alongside individual tool intents, there is a layer called “skills.” A skill is a Markdown file whose metadata helps Siri decide whether a request matches it. Siri reads the body of the file only after it has made a match.

That body can describe the action sequence, which intent should be preferred, what should be avoided, app-specific information, and when to ask a clarification question. In Apple’s shopping example, the skill coordinates the steps for product search, adding to cart, checkout and delivery method selection, reviewing the cart, and confirming the purchase.

Testing tools and today’s limits

The Siri Integration Advisor in Xcode provides feedback about the app. The Siri Evaluations framework makes it possible to script requests, measure responses, and catch differences as intents change. Developers can run signed development builds on supported devices and share them with internal TestFlight users.

The preview’s limitations are also clear. External TestFlight testers cannot use these capabilities. The preview also does not work in App Store versions of apps that include the new integration. Apple expects the first customer release within a few months, and App Store submissions are planned to open in the spring. Siri AI requires an Apple Intelligence-capable device and a supported language setting.

Similar capabilities are also spreading to Siri within Apple’s own apps: Calendar, Files, Mail, Maps, Messages, Photos, and Shortcuts. Apple’s examples include finding a café you’ve visited before and archiving files that haven’t been opened in more than a year.

Source: Unite.AI