Built for people who can't count on a signal
Your AI, With
No Signal Required.
The laptop runs the model. Pair your phone however you can reach it — WiFi, hotspot, mDNS, or nearby discovery. Same chat on both — no internet, no accounts, no cloud.
PortableAI
Local intelligence you actually own
Chat and search your own documents — with zero internet.
Run a local LLM on your laptop and ask questions over manuals, guides, and notes you load yourself. No API calls, no cloud index, no account.
- Default assistant works immediately
- Answers from files on your machine
Works entirely on your own hardware — no third-party services in the loop.
A vendored Ollama binary serves the model on the laptop. The iOS app pairs over a local WiFi hotspot or LAN with a PIN. No router, no internet, no cloud.
- Hotspot or LAN pairing
- PIN instead of an account
Add your own knowledge without editing files.
Create personas from the sidebar and attach local knowledge — recipes, field guides, procedures — so the assistant specializes when you want it to.
- Opt-in personas, never forced
- Your guides stay on device
Built for when the network is gone
Natural Language Processing
Local inference on your laptop — the model runs on hardware you hold, with no external API calls after setup.
Your files, your answers
Load your own manuals, guides, or notes — PortableAI answers from them, fully offline.
Personas, only if you want them
Chat works out of the box with a default assistant. Opt-in personas (Mentor, Explainer, survival-oriented roles, or your own) live in the sidebar.
Offline translation
Translate without a data connection — useful in the field when you cannot count on a signal.
Local pairing
The phone app connects to the laptop server with a PIN. No router, no internet, no cloud account.
Why carry PortableAI
Works with no signal
Full chat, files, personas, and translation with zero internet after setup. No cloud fallback in the critical path.
Your data stays yours
No user accounts. Each paired device only sees its own data. You can export it anytime. The laptop admin can review paired devices for storage and benchmarking.
Choose any model
Not locked to a fixed catalog. Run the GGUF or Ollama model you need — the bigger, laptop-hosted sibling to a smaller offline USB device.
Built for the worst case
Designed for crisis and off-grid use — someone alone, with no connectivity, who still needs an assistant.
Runs on hardware you own
A laptop plus your phone. Pair over a hotspot or LAN. No extra cloud subscription to keep it alive.
Intelligence off the grid
Sibling to USBMind (“AI. Off the grid.”). PortableAI is the version where you host the server and pick the model.
4 models tested, 0 refused real emergency questions
Fastest and most direct was survival-guide:latest at 8.4s average latency across the suite. Runs were on the same class of laptop the product ships with, not a cloud lab.
See the full benchmark →Get it running on your terms
Build Your Own
Download the repo and run PortableAI on a laptop you already have. No subscription. No account.
$0 one-time
$0 one-time
- Vendored local LLM server
- Default assistant, no account
- Phone pairs locally with a PIN
- Run any compatible model you provide
- You bring the machine and GPU
Questions people actually ask
Straight answers about PortableAI — no accounts, no cloud, no signal required.
Does it need internet?
What happens to my data?
Can I use my own model?
What if I don't have a laptop with a GPU?
Do I have to pick a persona to chat?
How does the phone connect?
Take intelligence with you when the signal dies
Self-hosted, fully offline AI for crisis, survival, and off-grid work. No internet. No accounts. No cloud.