AI on your Macs. With you anywhere.

Your Macs.
Your AI.
More together.

Run open models on the Macs you own. Chat from your iPhone, iPad, or browser with your protected Stratumus™ login.

Get early access Try the demo

Stratumus Connector for Mac · In testing

A weekend away
Plan a relaxed weekend in Lisbon.
Three days. Plenty of breathing room. Friday · Arrive, settle in, and wander Alfama. Saturday · Take the train to Sintra. Sunday · Pastries in Belém before heading home.
Everyday model · One Mac

Scripted demo · Context-based model routing is coming next.

The setup we’re building toward

One app.
Then you’re home.

Install the Connector. Connect your Macs. Access your local AI from anywhere.

Setup preview · The Connector is in testing. Automatic engine and starter-model installation are coming next.

Coming nextSmart routing

The right model.
Without you choosing.

We’re building routing that uses conversation context to choose an appropriate model. You describe the task; Stratumus handles the choice.

Plan a relaxed trip to Lisbon.

Planning a trip? Keep it light.

Everyday planning can use a smaller model on one Mac, leaving the others available.

Everyday modelOne Mac
Build me an app and test it.

Building an app? Bring everything.

More demanding coding could use a larger model shared through Mac Pool. Automatic routing and agent execution are future work.

Large modelMultiple Macs

The goal: fewer decisions.

Today you choose your model. Next, Stratumus will make the choice from what you’re working on, with manual control when you want it.

Proven in our testingMac Pool

Run bigger AI models.
Across your Macs.

Turn your connected Macs into a personal Mac AI cluster. Stratumus shares a model across their available memory, letting you run models too large for one Mac alone.

Llama 3.3 70B · 43 GB

Tap a Mac to add it or take it away.

Tested on real Macs. In our own testing, a 32-billion-parameter reasoning model too big for either of two Macs (16 GB and 24 GB) answered chats running across both. The interactive sizes above are illustrative estimates, not compatibility guarantees. What fits depends on available memory and model overhead; pooling does not guarantee faster replies.

In the app today Memory

It remembers.
You stay in charge.

Tell Stratumus what matters and keep it for next time. Memory is off until you turn it on. Saved memories live in your Stratumus account database; the AI runs on your Macs.

  • Opt-in. Off by default, and one switch turns it off.
  • Nothing hidden. See, edit or delete every memory.
  • Delete a chat, delete its memories.
  • Temporary chats use no saved memory and aren’t saved to your history.
9:41
Family
Remember that my daughter's name is Mia.
Memory saved
A week later
Write a birthday note for Mia.
Message
From anywhere

Your login.
Your Macs. Your AI.

The open-source models run on your own hardware, so there are no per-request AI fees.

Reach home from anywhere, safely.

  • Nothing to open on your router. Your Macs connect out; no port forwarding, no VPN.
  • Approve each Mac once from your phone, and revoke any Mac instantly.
  • Encrypted in transit. The AI runs on your Mac, not in someone else's cloud.

Choose from supported open models, ranked for each Mac.

LlamaQwenGemmaDeepSeekMistralgpt-ossPhi
FAQ

Questions, answered.

What works today?+

Pairing your Macs, chatting with them from any device, switching and downloading models remotely, memory, and Mac Pool (sharing one model across multiple Macs). The Mac app is in final testing, and today it still expects Ollama or oMLX to be installed first. Setting everything up for you, and smart routing, are what we're building next.

Do my chats leave my home?+

Model inference runs on your Macs. Messages pass through the Stratumus service, encrypted in transit, so you can reach them remotely. Conversations and saved memories are stored in your Stratumus account database. This is local inference with a hosted account service. Temporary chats aren’t saved to your history.

Where does memory live?+

In the Stratumus account database, not on your Mac. Memory is opt-in, and you can view, edit or delete saved items. Deleting a chat deletes the memories made in it.

Does smart routing work today?+

Not yet. Today you pick the model on each Mac. Choosing automatically from the conversation is next. The demo above is a preview.

How does Mac memory pooling work?+

Stratumus splits one supported model across the available memory of multiple Apple silicon Macs on your home network. This is distributed LLM inference: each Mac handles part of the model. It expands model capacity without turning the Macs into one shared system-wide RAM space.

Will Mac Pool make my models faster?+

Mac Pool expands the models your Macs can fit together. It doesn’t guarantee faster replies or better answers. Speed depends on the model, your Macs, and the connection between them.

What do I need?+

A Mac with Apple silicon and a supported local model engine. The Connector currently expects Ollama or oMLX to be installed first. Mac Pool needs at least two compatible Macs connected to the same local network. To answer while you’re away, your Macs need to stay awake and connected.

Does it work with Windows PCs?+

Not yet. We're making Stratumus excellent on Mac, iPhone and iPad first.

Your Macs.
More together.

Stratumus is coming to macOS in late 2026. Leave your email and we'll tell you when it's ready.

One email when it's ready. No spam. Privacy