About
Private AI that works for you — on your own Mac
Cloak Forge sets up private AI automation on your own Apple Silicon Mac — or, if you’d rather not set anything up, runs your data through our computer for you. Either way, nothing is sent to a cloud service, so there’s nothing to leak, nothing to meter, and nothing that stops working when a vendor changes its terms.
What we do
Private AI — on your machine or ours
Build it for your machine
We pick a model sized to your Mac’s chip and memory, build and test everything on our end, then send it to you ready to run — no accounts, no API keys, no monthly plan.
Automate the busywork
We set the model up to work with your kind of files and tools — sorting and drafting email, summarizing documents, organizing data — and schedule it to run on its own.
Teach you to run it
Prefer to do it yourself? We’ll teach you how to choose models, build pipelines, and fix things when they break — on your hardware, at your pace.
Or run it for you
No setup on your end at all: send us your documents or data and we run them through the local AI on our own computer — never a cloud service — then send you the results.
We never store or sell your data.
Why local
Why run AI on your own computer?
Your data stays yours
Your files are only ever processed on a physical machine — yours, or ours — never on a cloud AI service. There’s no third party holding copies, and no vendor account to be breached.
No meter running
Cloud AI bills per word or per hour, forever. Local AI costs the setup and the electricity your computer already uses.
Works offline
It keeps running when the internet drops, when a provider changes its prices, or when a cloud product shuts down.
Fit check
Will it run on your Mac?
We only build for Apple Silicon Macs — any Mac with an M1, M2, M3, M4 or newer chip. Their unified memory lets the AI use nearly all of the computer’s memory, which is what makes big, capable models practical on a desk. How much memory your Mac has decides which models it can run — find it below, then use the checker.
Step 1: find your chip and memory
- Click the Apple menu in the top-left corner of the screen.
- Choose
About This Mac. - Read the Chip line. If it says Apple M1, M2, M3, M4 or newer (any variant, like “M2 Pro”), you’re set. If it says Intel, see the note below.
- Read the Memory line — e.g. “16 GB.” That whole number is usable by the AI.
Step 2: check your fit
Intel Mac, Windows or Linux? We don’t build for those — PCs split memory between the computer and the graphics card, and Intel Macs are too slow for these models, so neither can match an M-series Mac without expensive extra hardware. You can still use us: we’ll run the job on our own Mac and send you the results.
Memory guide
What to expect at each amount of memory
Memory on an Apple Silicon Mac. The figures below assume 4-bit quantized models — the format most people run locally — with room left for a working context window.
| Memory | Model class that fits | What to expect |
|---|---|---|
| 8 GB | 3B – 4B | Basically unusable Only tiny models fit, and they get things wrong too often to rely on. We don’t recommend AI work on an 8 GB Mac. |
| 16 GB | 7B – 8B | Only slightly usable Fine for short, simple jobs — quick summaries, sorting, tagging — but it struggles with anything longer or more complex. 24 GB is where it starts to get comfortable. |
| 24 – 32 GB | 12B – 14B | A clear step up in reasoning and instruction-following, or an 8B model with a long context window — good for daily work. |
| 48 GB | 27B – 32B | Strong on nuanced writing, code, and multi-step tasks. Close to mainstream cloud quality for most jobs. |
| 64 GB | 70B (4-bit) | Where local output starts to rival premium cloud models, with roughly 40 GB of weights and headroom for a large context on top. |
| 128 GB + | 70B at full precision, or large mixture-of-experts models | Room to keep several models loaded at once and hold long-running context in memory — the basis for always-on agents. |
See what I actually ran on four Macs, from 16GB to 48GB: How much RAM you need to run local AI on a Mac.
Why we only work on M-series Macs
Unified memory changes the math
On an Apple Silicon Mac, the chip and its graphics share one pool of memory, so the AI can use nearly all of it. A 64 GB Mac runs models that would otherwise need a data-center graphics card — and a Mac around $2,000 covers most real workloads.
On Windows and Linux PCs, the model has to fit inside the graphics card’s own memory, which tops out around 24 GB on consumer cards. Going bigger means $2,000+ cards or very slow processing. Focusing on one platform also means every setup we ship is one we’ve built and tested on the same kind of machine you have.
When you get in touch, we look at what you already own, tell you plainly which tier you’re in, and lay out what it will and won’t do before you spend anything on hardware.
What to expect
From first message to finished setup
-
You reach out
Tell us what you want to automate. We reply within one business day.
-
A 20-minute call
We talk through what you run now, what hardware you have, and which option fits. No sales script.
-
We build it
We build and test everything on our end, then send it to you ready to run — or we run the job on our machine and send back the results.
-
It’s yours
You keep full access and we keep none. No subscription, and no copies of your data on our side.