Feature request: background input + Accessibility API for Coherence windows (AI agents / automation)

AlexR44

Bit poster
Hi all,
I'd like to request (and gather +1s for) a long-standing gap in Coherence: macOS automation tools can see Coherence windows, but they can't operate them unless they take over the whole screen.
Setup
  • Parallels Desktop 27.0.3 (build 58680), Standard Edition, on macOS 26.6.2 (25G83)
  • Windows guest, app running in Coherence
  • Windows-only business desktop app (an enterprise planning/admin console)
What happens
  • Screenshotting a Coherence window works fine, even when it's in the background.
  • Mouse events sent to the Coherence window from macOS arrive (the pointer visibly moves onto the target), but the Windows app never responds. I tried ribbon tabs, document tabs and buttons, and nothing registered.
  • The window's macOS accessibility tree is completely empty, so tools have no buttons, tabs or fields to target.
  • Input works only when the window is brought to the front and driven as the active app. For an automation tool, that means controlling the entire display.
What I'm asking for
  1. Forward background input. Pass mouse and keyboard events aimed at a specific Coherence window into the guest app even when the window isn't active, the way native Mac apps receive them.
  2. Accessibility bridge. Map the guest app's Windows UI Automation tree (buttons, tabs, grids, text fields) onto the macOS Accessibility API for Coherence windows.
  3. Opt-in. If security is a concern, put both behind a per-VM or per-app setting that's off by default.
Why it matters
  • AI agents now operate Mac apps. AI assistants and automation tools increasingly work desktop apps for the user through the macOS Accessibility API and background input. Native Mac apps support this, but Windows apps in Coherence don't. For anyone whose key software is Windows-only, Parallels becomes the blocker.
  • Least privilege. Today the only working option is to give a tool control of the whole Mac screen. Per-app input would let you grant access to one Windows app and nothing else, which is better security and easier to get past IT.
  • You keep working. Background control lets the tool work in one app while you use others. Full-screen takeover locks you out until it finishes.
  • Accessibility. The same bridge would make Windows apps in Coherence usable with VoiceOver and other macOS assistive technologies.
  • It completes the Coherence promise. Coherence makes Windows apps look like Mac apps. They should also behave like Mac apps when tools interact with them.
This has come up before in different forms: a 2010 request for Accessibility API support on Coherence windows (a Parallels developer replied "We plan to support Accessibility API in the near future"), a 2008 question about voice-driving Coherence apps with macOS assistive tools, a 2013 thread on Mac window managers not working with Coherence, and a 2023 report that Keyboard Maestro never works in Windows apps. They all trace back to the same gap: Coherence windows look like Mac windows but don't take part in macOS accessibility or input the way Mac windows do. AI agents that operate apps make this much more pressing now.
If you'd use this too, please reply with your use case. Parallels team, I'm happy to provide repro steps or test a beta.
Thanks!
 
Back
Top