The Mac at home,
in your pocket.
Pallm turns the open models on your Mac into an assistant you carry everywhere. Chat, search the web, make images and send coding agents to work, all on hardware you own.
No account and no cloud. Your chats live on your Mac.
For macOS 15 or later, Apple silicon. The iPhone app needs iOS 18 or later.
Works with
- LM Studio
- Ollama
- oMLX
- llama.cpp
- mlx-serve
- ComfyUI
- Draw Things
A reply that finds its way back
Most local AI apps only work while you stare at them on your home Wi-Fi. Pallm was built for the walk to the bus.
Your desktop's models, on your laptop
Leave the big Mac at home. Pallm on your MacBook connects to it from anywhere, on hotel Wi-Fi or your phone's hotspot, and brings every model it serves with it.
- The big model, on the small Mac A MacBook Air with 16 GB gets the 120-billion-parameter model your Mac Studio runs, at the Studio's speed, and the battery barely notices.
- The same chats, wherever you open them Start on your phone on the way in and carry on at your desk. A reply that's still being written streams on the laptop from where it is.
- Nothing to run on the laptop It's the same Pallm app, set up as a client: no models to download, no service in the background, no menu bar icon.
- Your coding tools come too Switch on the local model gateway and pi, opencode or any OpenAI-compatible tool on the laptop uses the model at home. No API key from anyone.
What you'd use a cloud assistant for, at home
Everything runs on your Mac, against whichever model you pick. Your phone is the remote control.
- The servers you already run LM Studio, Ollama, oMLX and llama.cpp on your Mac are found by themselves. Pallm loads and ejects models to fit your memory.
- Search and read the web Chats look things up and read pages for you, through DuckDuckGo or your own SearXNG. No API key needed.
- A browser it can drive For pages that need clicks, the model works a private Chrome on your Mac while you watch it in the side panel.
- Memory, if you want it Off until you turn it on. Then it keeps a short list of what's worth knowing about you, which you can read and correct.
- Projects Chats that share instructions, notes and files. Whatever a chat makes lands on the project's shelf for the next one.
- Artifacts that run Web pages, React components, charts and diagrams run right in the chat, offline, on your phone.
- Agents that open pull requests Hand a coding agent a repo and a task. It works in a sandbox on your Mac and pushes nothing until you approve the commit.
- Scheduled tasks A prompt that runs every weekday at nine, or every fifteen minutes, and only notifies you when it finds something.
- Images, music and video Ask in a chat, and ComfyUI, Draw Things, mlx-serve or ACE-Step make it on your Mac. In beta.
- Skills Folders of instructions in the same format as Claude's skills. Import one from GitHub and it works as it is.
- Your calendar, when you allow it Chats can check your calendar, add reminders and read Health data. Each one asks first, or stays off.
- One Mac, the whole household Everyone who shares the Mac gets their own chats, memory and settings, with an optional password.
Nobody in the middle
Your chats live on your Mac. There's no Pallm server to sign in to, and nothing we could read if we wanted to.
- Encrypted all the way. Your phone and Mac talk over WireGuard. Each device has its own key, and one you remove can't get back in.
- Paired with a QR code. Scan it from your Mac's screen. That's the whole sign-up.
- Relays see only envelopes. When a direct path isn't possible, a relay passes encrypted traffic along. It sees that two devices talk, never what they say.
- Notifications stay vague. They say "Reply ready", never what the reply says.
Set up in about ten minutes
- 1 Install Pallm on your Mac Open it and choose Set Up This Mac as the Host. It keeps running in the menu bar.
- 2 Give it a model Pallm finds LM Studio, Ollama, oMLX or llama.cpp by itself. New to local models? Start with LM Studio.
- 3 Scan the code with your iPhone Pallm on the phone pairs with your Mac, and you're chatting.
Using a MacBook as the host? Listen mode keeps it awake with the lid closed while it's plugged in.