Apple Engineers In-House AI Agent to Execute Cross-App Workflows

A recently removed job listing reveals Apple is developing an in-house AI agent capable of controlling desktop apps, writing code, and orchestrating sub-agents across macOS and iOS.

AI & Culture Desk
2 min read
Apple Engineers In-House AI Agent to Execute Cross-App Workflows

Apple is building an in-house AI agent capable of operating applications, executing command-line tasks, writing code, and navigating the web on behalf of users. According to reports from Bloomberg, The Verge, and TechCrunch, details of the project surfaced through a now-removed job listing for a senior engineering manager to lead the agentic AI assistant team. The development signals Apple’s intention to compete directly with established agentic systems from OpenAI, Anthropic, Meta, and Google, while leveraging its integrated operating systems to deliver deep cross-application automation.

The role’s job description specifies that the system will execute multi-step workflows across software tools, web browsers, command lines, code editors, and APIs. To accomplish complex instructions, the central agent will coordinate specialized sub-agents while adhering to Apple's privacy standards. Engineering for the project spans three key groups within the company: the Siri team, UI Frameworks, and Apple Foundation Models. According to Bloomberg, progress remains in its early stages, though Apple has considered release windows as early as late next year.

The initiative builds upon Apple's incremental rollout of AI features across its platforms. In September, the company introduced an upgraded version of Siri designed to understand on-screen context and execute actions within apps. Upcoming system software updates, including iOS 27.3, will expand developer access through a Developer Preview that allows external applications to integrate with Siri. Bloomberg reports that native agent capabilities will eventually handle specific multi-step tasks like organizing tabs in Safari or cropping and rotating images within the Photos app.

Apple’s push into autonomous control coincides with tighter platform restrictions on competing software. As reported by TechCrunch, upcoming builds of macOS 27.3 will deploy explicit warnings to discourage users from granting full system access to third-party AI agents. In October, Apple notified developers of plans to add controls ensuring that users only grant this level of access through very explicit actions. While competitors such as Meta’s Muse, OpenAI’s Dots, and Anthropic’s Duet offer desktop control tools, Apple is placing structural hurdles in front of third-party agents while developing its native alternative.

By pairing strict system security with native sub-agent orchestration, Apple is leveraging its control over OS-level APIs, accessibility frameworks, and system hardware. Platform vendors maintain an inherent advantage in desktop automation because third-party agents must navigate permission boundaries and visual screen scraping. If Apple delivers low-level system hooks tied to its foundation models, it could offer higher execution reliability and tighter privacy controls than cross-platform competitors running on top of macOS.

Earlier in this story: Apple Tightens macOS Disk Permissions to Counter AI Agent Risks (Oct 3, 2026)

Sources

  1. 01 Apple is hiring engineers to build its own AI agent — The Verge — AI
  2. 02 Apple may be developing its own AI agent that runs your Mac — TechCrunch — AI