Izuran on their Macs
The assistant app, linked to your server instead of the cloud.
Private AI Server · Powered by Izuran Pro
The Private Mac AI Server Kit
Everything stays in the box.
A private AI server built on Mac Studio or Mac mini. It runs open-weight models on your own network, with agent workflows that are governed, logged and reviewable.
What it is
A Mac Studio or Mac mini, or two Mac Studios joined by a Thunderbolt 5 cable, running Izuran Pro, our own server software. It serves open-weight models to your team on your network and runs the agents that do the work.
How your team reaches it
Izuran for iPhone and Apple Watch
Text or voice, straight to the server over your network or your Tailscale network. No relay service.
The tools you already use
Through an OpenAI-compatible API on your network, over TLS, with a key per client.
What stays in the room
Your prompts, documents and conversations are processed and stored on the server in your office. The models run there too. The server only connects outside your network for the reasons below, and it keeps a network log your team can read.
By default
- Signed software updates
- Model downloads, only for models we've tested. Your server arrives with its models installed.
- Apple's push service, only if an iPhone is paired
Off until you turn it on
- Reading web pages
- Web search, sending only the search words
- Gmail, Signal, and Apple Weather, Maps and Music
- Connections to other tools you add
- Remote support from AIC, through your Tailscale network, only with your approval
Never
- A cloud AI model
- Third-party telemetry
- A relay service between your phone and the server
Offline switch. One setting limits the server to your own Macs and devices, update checks included. Your team can open the network log any time to see what connected where.
What's in the kit
Three sizes, the same software. Memory decides which models fit, so we start from your use case, not a spec sheet.
The largest models run split across two Mac Studios working as one cluster appliance. It's all headless: no monitor or keyboard needed after setup.
- Compact
- Mac mini with M5 Pro, 64 GB, for smaller models
- Studio
- Mac Studio with M3 Ultra, 512 GB
- Cluster
- Two Mac Studios, 1 TB combined, as one appliance
- Software
- Izuran Pro on every size, models chosen to fit
- In the box
- Rack with a built-in status display; Thunderbolt 5 cable for the cluster
- Your devices
- Macs and iPhones link to it; other tools use the API
Models by size
- Mac mini, 64 GB
- Qwen 3.8, Gemma 4, Muse Glimmer
- Mac Studio, 512 GB
- All of the above, plus Qwen 3.8 Flash-Next and GLM 5.3 Flash
- Two Mac Studios
- All of the above, plus GLM-5.3 and DeepSeek V4.1 Flash, split across both
Governed agents
Every agent is governed like a new employee. It works in the folders you give it, and changes outside them need approval. Anything risky waits for a person, and the riskiest actions need Touch ID or a password. Every conversation is logged on the server.
How the controls work →Setup and support
It's plug and play. The server arrives with the latest Izuran Pro and your models already installed. We stay up to two hours to set it up with you and walk your team through it.
Support is by phone, email or in person, and remotely through your Tailscale network when you approve it. Software updates install themselves, checked against our signature, and a failed update rolls back.
Pricing
Pricing depends on configuration and support. We'll go through it on the qualification call.
See it in your office first.
A request doesn't guarantee a trial. Trials are limited, available only in our delivery area, and offered when the use case is a fit.