Apple please approve!

[cpietsch]

we need this!
thank you for creating this @beriberikix

[wheaney]

I don't suppose there's some kind of ticket we can upvote to show support for this, huh? Would reaching out to Apple help?

[beriberikix]

An update, since this has been open a while and the situation is now clearer than "waiting on Apple".

The request was declined, on 25 February 2026. The Developer portal shows it plainly:

Capability	Status
DriverKit USB Transport - VendorID	Declined
DriverKit UserClient Access	Declined

The denial redirected to com.apple.developer.usb.host-controller-interface, which unfortunately does not address this project's problem — that API is for implementing a USB host controller, which is what a USB/IP client needs (see #7), not what a server needs to take a device from its driver.

There is no userspace workaround, and that is now measured rather than assumed. Against real hardware — a keyboard, a flash drive and a webcam:

• USBInterfaceOpen → kIOReturnExclusiveAccess
• USBInterfaceOpenSeize → kIOReturnExclusiveAccess, identical. Seizing does not displace a kernel driver.
• Unmounting a disk → still blocked; the filesystem detaches but IOUSBMassStorageDriver stays bound
• Ejecting it → still blocked; the media stack drops from 20 IOKit nodes to 7 and the interface remains owned

So the blocked device classes are genuinely blocked. Anyone hoping a clever flag would unlock USB-serial adapters can stop looking.

A narrower request is in flight. Request ID 26F53XCAGY, submitted 2026-08-06, for DriverKit Family Serial at the development tier only — a much smaller ask than the vendor-wide production entitlement that was declined, and enough to prove whether a DriverKit extension can claim a USB-serial adapter at all. That question has never been testable, because AMFI kills a binary carrying restricted entitlements without a matching provisioning profile.

Meanwhile the project works for what it can serve. v0.5.1 is out and verified against hardware: a J-Link driven end to end from a Linux client with probe-rs, and an Android device in ADB mode answering protocol messages over the wire. Devices macOS does not claim need no entitlement and never did. bind now refuses the rest up front and names the owning driver, instead of accepting them and failing later at transfer time.

Details in Documentation/development/entitlement-validation.md and driver-free-release.md.