YgritteReading context
Research guide
Files, notes, and source checks
Available on iOS, Android, Mac, Windows, and Linux
Stop renting AI.
Own it.
PocketAI turns the hardware you already own into your AI engine. Download a model once, then chat, create, and run agents on your own machines — no per-token bills, no monthly chat subscription, no cloud in the middle.
Runs on your hardwareWorks offlineAgents with approvalsPay once, use forever
One app, four doors
Built for the way you use AI
The same local-first engine underneath. Pick the door that sounds like you.
Developers
An OpenAI-compatible API on localhost, a CLI bridge for your coding agents, and a VS Code extension — all powered by models you control.
Point your tools at localhostCreators
Generate images, video, and 3D assets from one workspace, organized into projects, with the heavy lifting on your own GPU.
Create without a render billOffline & mobile
Download a model once and chat anywhere — airplane, subway, dead zone. Conversations stay on the device they happened on.
Take your AI off-gridIdle hardware owners
That gaming PC doing nothing all day? Make it your personal AI endpoint and reach it securely from your phone or laptop.
Put your GPU to workFor developers
Your local models, behind an API you already know
Start a local server from PocketAI and every installed model is available at an OpenAI-compatible endpoint. Existing SDKs, scripts, and tools work by changing one base URL — and nothing you send ever leaves your machine.
OpenAI-compatible /v1/chat/completions, /v1/completions, and /v1/models on localhost
CLI bridge for Codex, Claude, OpenCode, Cursor, and custom command runners
VS Code extension wired to your local server out of the box
Install multiple models, switch per task, import your own files
terminal
$ curl http://127.0.0.1:39457/v1/chat/completions \
-d '{"model": "local", "messages": [...]}'
{
"choices": [ ... ],
"served_by": "your own machine"
}Create workspace
No credits usedRunning on your GPU
Image
Video
3D mesh
Projects
Generate
For creators
Images, video, and 3D — one workspace, zero meter
The Create workspace runs image, video, and 3D mesh generation on your own hardware where it's supported. Iterate as many times as you want; the only cost is electricity.
Image, video, and 3D mesh generation from a single app
Histories and project folders for everything you generate
Per-mode generation settings you can actually tweak
Kick off heavy renders on your desktop from your phone
Offline & mobile
Download once. Chat anywhere.
Once a model is on your phone or laptop, PocketAI needs no connection at all. Voice conversations, file and image attachments, and your full chat history — all living on the device in your hand.
Truly offline
No internet needed after the model download
Native apps
iOS and Android, with device-fit model picks
Voice mode
Local speech-to-text and text-to-speech
For idle hardware
Your desktop GPU, reachable from anywhere
Claim your desktop, connect it to your personal PocketAI subdomain, and use its models from your phone, laptop, or browser. Big-model quality on the go, served by hardware you already paid for.
Personal subdomain for each claimed desktop
List and load the desktop's models from any signed-in device
End-to-end encrypted remote traffic between your devices
Run desktop agents remotely and approve sensitive steps from your phone
Devices
2 devicesYour personal AI endpoints
Gaming PC
you.pocketaihub.com
Your phone
Using the PC's 27B model
End-to-end encrypted between your devices
Your agents
2 activeWorking on your desktop
MagnusNeeds approval
Automation operator
Shell, browser, and app control
Approve next step
Agents
A team that works your computer, not a cloud
PocketAI agents run where your files, browser, shell, and apps already live. Give each one a role, a brain, and its own context, then queue up work and supervise it from your phone — approving the sensitive steps yourself.
Agent profiles with roles, brains, teams, tools, and memories
Queue, monitor, and review agent work items in one place
Approve risky actions from your phone; pause, resume, or stop any time
Power agents with local models or bridged CLI brains like Claude and Codex
PocketAI Compute
Coming soonNo GPU? Rent one by the minute.
When a job needs more power than your hardware has, spin up a cloud GPU worker that plugs straight into the same PocketAI workspace — same models, same OpenAI-compatible endpoints, same agents. It stops itself when idle, so you only pay while it's actually working.
GPU workers provisioned on demand for heavier models
Remote agent workspaces backed by cloud machines
OpenAI-compatible endpoints, just like your local server
Automatic idle stop — no forgotten meters running overnight
Privacy
Private by architecture, not by promise
We don't need a privacy policy full of exceptions, because the data never arrives. Local-first isn't a feature toggle — it's how PocketAI is built.
On-device by default
Conversations are stored where they happen
Encrypted in transit
Remote access is end-to-end encrypted between your devices
Nothing to leak
No cloud chat logs means no cloud chat breaches
Pricing
Own the engine. Rent nothing.
Local features are yours forever with one payment. Pro only exists for the network layer — reaching your machines from everywhere else.
Free
$0forever
A real local AI starting point. Try the core workflow on your own device before spending anything.
DownloadLifetime Local
$39.99one-time
Own the local engine: full model use, desktop and mobile workflows, local creation, and local agents. Forever.
See what's includedPro
$4.99per month
Unlock the network layer: remote access, your personal subdomain, remote agents, and multi-device workflows.
See what's includedYour hardware is ready
Install PocketAI, download a model that fits your device, and have a private assistant running in minutes.
Download PocketAIEarn with PocketAI
Recommend PocketAI and earn 30% on every lifetime purchase you refer.
Join the affiliate program