Your phone chats with AI models on your Mac

One command on the Mac serves its Ollama models over your Wi-Fi and prints a QR code. Scan it: your phone is the screen, the Mac runs the model.

npx github:VedSoni-dev/berth serve

Open the dashboard

Free and MIT licensed. The Mac needs Node 18 or newer and Ollama.

What the phone app does. Your Mac serves it, so there is nothing to install from a store.

Model list: every model on your Mac

Tap the model's name to pick any model Ollama has on the Mac. One already in memory answers with no wait to load.

Speed from the Mac, under every answer

Every answer ends with its speed, the time to the first token and how many tokens it took.

Home Screen icon on your phone

Add it from the browser's Share menu and it opens full screen with berth's icon, in light or dark.

Pairing code between phone and Mac

The QR code carries a pairing code, and the Mac answers nothing without it. berth serve --new-token unpairs every phone.

Set up berth in 3 steps

  1. Run one command on your Mac

    Paste it into Terminal. berth finds Ollama, then prints a QR code and the address to open.

  2. Scan the QR code with your phone

    Use the camera on the same Wi-Fi. The chat opens with the pairing code already in its address.

  3. Ask the model on your Mac

    Type a question. The answer streams in as the Mac writes it, with its speed underneath.

The dashboard runs a model in a tab

No Mac at hand? Open the dashboard in Chrome or Edge and run a small model on this computer's GPU, or connect Ollama.

  • In this tab

    WebGPU runs Qwen, Llama, Gemma, Phi or SmolLM, with nothing to install.

  • Ollama

    A local OpenAI-compatible endpoint, model pulls and a second machine.

  • Your phone

    The command to run, and whether a phone is paired.

Open the dashboard

Questions

What berth needs, what it can't do yet, and why it costs nothing.

berth is free on publik, a free software marketplace. When you use it through publik, its builder earns a small share of the usage, which keeps berth free. publik's mission: free software for everyone.

Make a free publik account →

Tailscale, for your phone away from home

berth has no relay of its own. Put the Mac and the phone on Tailscale, then open the Mac's Tailscale address with :7770. The Mac has to be awake.

The Mac: Node 18 and Ollama

Node 18 or newer, and Ollama with a model pulled, such as ollama pull llama3.2. berth adds no dependencies.

Our servers: none between phone and Mac

No. The phone app, the proxy to Ollama and the pairing check all run on your Mac, over your Wi-Fi.

Not done yet: pooling memory across Macs

Pooling memory across machines to run one bigger model, a menu-bar app, and a relay. Each Mac runs whole models.

Android phones and .local addresses

Many Android phones can't open .local names. Run berth serve --ip and scan the new code.