I built a web API that lets you access Apple Intelligence from any device or programming language.
It’s usually locked to Apple silicon and can only be accessed in Swift.
By building a web server that mimics the OpenAI/OpenRouter API format, I make it possible to call Apple Foundation models from Windows, Linux, Python, JavaScript, or anywhere else.
I walk through why you’d want this, implementing the API in Swift using Vapor, making it compatible with OpenAI’s standard, and stress testing it on my M2 MacBook Air.
GitHub repos:
– Main API: https://github.com/gouwsxander/Apple-Intelligence-API
– Load tester: https://github.com/gouwsxander/stress-test
– Minimal chat interface: https://github.com/gouwsxander/minimal-chat
Links to things I mention or found useful:
– Apple Foundation Models docs: https://developer.apple.com/documentation/FoundationModels
– Foundation Models deep dive (WWDC25): https://youtu.be/6Wgg7DIY29E
– Vapor web framework: https://vapor.codes
– OpenRouter API: https://openrouter.ai/docs
– CNET on Apple Intelligence: https://youtu.be/D0V554NyXWM
– MKBHD on Apple Intelligence: https://youtu.be/hz6oys4Eem4
– The Verge on Apple Intelligence: https://youtu.be/ghS7qcBK0Gw
Timestamps:
– 0:00 Introduction
– 1:29 The solution
– 2:20 Development
– 3:34 Trying it out
– 4:45 Future improvements
– 5:20 Unicorn startup?
Note: Xander Gouws is the original author of this video, we just embed it, if you have any questions please contact him via Youtube.
