Private AI that runs on your Mac
Local: AI that runs on your machine.
No cloud, no accounts, zero $ per token.
Privacy Mode
- Nothing leaves your device. Every request runs on your Mac.
- Memories stay on the machine, never sent to a cloud provider.
No data leaving your device
The more memory,
the more Local can do
Zero setup, it just works.
A whole workspace,
none of it in the cloud
Chat with your files
Ask questions, draft writing, drop in a PDF and get it summarised.
Code in your folder
Point it at a project and it reads, edits and runs code in place. Your codebase is never uploaded.
Meetings, written up
Transcribed on device, with speakers labelled.
Memory you control
A short list of lasting facts about you, all of it visible and editable. Mark one sensitive and it never leaves the Mac.
Models that fit
Local reads your chip and memory, then recommends what to run.
See what stayed
Tokens, speed per model, and what the same work would have cost through a cloud API.
Office Mode
One device does the work, the office shares it. Run the big model on the most capable machine you own and reach it from Local on every laptop in the room.
NVIDIA DGX Spark · On-prem Server
Open models, free and ready to run
Qwen
Llama
Gemma
Mistral
Phi
DeepSeek
WhisperMore intelligence only when you need it
Some jobs want a frontier model.
Connect one with your own key, then go back to local for everything else.
A provider in your country
Route to a neocloud that keeps inference inside your jurisdiction.
More regions being added
The frontier labs
Bring your own key and their top models join the picker beside your local ones.
OpenAI
Anthropic
OpenRouterIt tunes itself to your chip
Local ships BaseRT, our inference engine, compiled for your silicon on first launch. Same model, same Mac, more tokens per second.
Get Local for your machine
We built Local because powerful AI should feel personal, not like renting time on someone else's server.
Does it work without internet?
Yes. You need a connection to download the app and a model. After that, everything runs offline.
Which Macs are supported?
Any Mac with Apple silicon. Local picks a model that fits your memory.
Is any of my data collected?
No. Everything lives in a database on your Mac. No account, no telemetry.
Can I still use GPT or Claude?
Yes, with your own API key. Those requests go to the provider, so cloud models are off while privacy mode is on.
How does my machine affect performance?
Memory decides which models fit, so Local gets more intelligent the better your machine. More unified memory runs larger, smarter models; a faster chip runs them quicker. Local reads both and picks the best setup automatically.
What makes it faster than other local apps?
BaseRT, our inference engine. Most local apps ship one generic build. Local installs the one compiled for your chip.
