Your Mac or PC as your iPhone's private AI backend. Install, click Share with my phone, scan the QR in MyLLM → chat with your own models over a trusted, encrypted link — at home or anywhere. Nothing transits any cloud.
Downloads
| Platform | File | Notes |
|---|---|---|
| macOS 10.15+ | MyLLM.Connect_0.1.0_universal.dmg |
Universal (Apple Silicon + Intel) · signed & notarized — no Gatekeeper warning |
| Windows 10+ | MyLLM.Connect_0.1.0_x64-setup.exe |
Not code-signed yet — SmartScreen will warn (More info → Run anyway); signing is on the roadmap |
What it does
- First-run setup: checks your model (guides Ollama install / model pull:
llama3.2,qwen2.5,gemma2:2b) → secure link → pairing QR. - The lock: a loopback-only bearer-auth proxy in front of Ollama — 401 without the paired key; NDJSON chat streams straight through. Rotate the key anytime to cut off all paired devices.
- Trusted HTTPS anywhere:
tailscale serveover your own private mesh — valid certificate, no port forwarding, nothing public. - Menubar/tray native: template icon + menubar-only on macOS; system tray on Windows.
Requires
Ollama and a free Tailscale account — the app guides you through both.
Protocol & source are in this repo (Apache-2.0): see PAIRING_PROTOCOL.md.