On-device · nothing leaves your machine

Your Private On-device AI Assistant that gets stuff done.

One private assistant for everything you do. Control your computer by with new modalities, Create with AI that writes in context, and Automate entire workflows — all without sending your data to the cloud.

  • Control
  • Create
  • Automate
  • Extend
SCROLL TO WATCH IT WORK

Auto-playing product animation for Cephable, an on-device assistant. It has no interactive controls — the same information is described here, in three stages.

Stage 1, Control. Cephable runs entirely on your device and listens for voice commands across the whole computer. In the demo it opens Outlook, starts a new email, dictates the message, and sends it — the whole task completed hands-free. Nothing leaves the machine.

Stage 2, Create. Describe what you want to write in plain language and Cephable drafts it in place — here it turns meeting notes into a full product requirements document. The same generate, rewrite, and translate tools work inside any app.

Stage 3, Automate. Hand off a multi-step task in a single sentence. Cephable carries it out across your apps — in the demo it drafts a follow-up email and updates a Salesforce CRM record from a call transcript — then leaves everything for you to review.

AND HERE’S THE PART NO CLOUD AI CAN CLAIM

All of it runs right here.

Not in a data center. Not on someone else’s servers. On the device in front of you. Keep scrolling — we’ll take you all the way down to the silicon.

Cephable runs entirely on your device. It starts as the app on your laptop, goes past the display into the motherboard, and all the way down to the silicon chip doing the work. The AI model runs on your own hardware — private by design. Nothing ever leaves your machine. No cloud. No data collection. Works offline.

It all runs right here.

Every command, every word, every workflow — processed entirely on your device. Private by design. Nothing ever leaves your machine.

No cloudNo data collectionWorks offline
GET THE MOST OUT OF THE COMPUTER YOU'VE INVESTED IN

Built for all computers.
Optimized for the AI PC era.
Validated by the makers.

Intel Qualcomm AMD NVIDIA Lenovo HP Microsoft Mac

Cephable runs on machines with a variety of compute levels and activates value on your device: turning hardware capability into a concrete local AI experience.

SCROLL TO SEE WHAT TO EXPECT

Tell us about your machine and see exactly what to expect — which model you'll run and which plan fits, entirely on device.

On-device · nothing leaves your machine

Your machine

Rough is fine
Recommended for you Cephable Gemma M The everyday assistant
4B parameters · 16 GB RAM · ~16K token context
Insight level

A light sounding board

Handles brainstorming and short reasoning on device. Deeper work routes up to a larger model, and you'll always see when it does.

Recommended plan

What you can do

An estimate based on what you told us. Once installed, Cephable reads your actual hardware and tunes itself, so your real experience may land a notch higher or lower.

HOW IT WORKS

On-device from the first word

No accounts to wire up, no prompts shipped to a server. Cephable installs a private AI model that lives on your machine and gets to work.

01

Install and it's yours

Download Cephable and a private AI model runs locally. No subscription required to start, no cloud round-trips.

02

Use with any app

Cephable operates as your assistant in any app you have installed. Anywhere your cursor goes is available to Cephable.

03

Control, create, automate

Use Cephable in the ways that mean the most to you. Controlling multi-modally, creating content or automating workflows

DO MORE THE WAY YOU WANT

One assistant. Endless reach.

No accounts to wire up, no prompts shipped to a server. Cephable installs a private AI model that lives on your machine and gets to work.

Control
⌘

Voice control

Drive your entire computer hands-free. Open apps, click, switch windows, and trigger any action with your voice.

Control

Dictation anywhere

Speak and your words flow into any text field — email, docs, chat, code. Punctuation and formatting handled for you.

Control

Custom commands & macros

Map a phrase to anything — keystrokes, app actions, shortcuts, multi-step macros. Build the controls you wish existed.

Create

Generate text

Describe what you need and Cephable writes it where you're working — emails, docs, replies, summaries, and more.

Create

Rewrite & translate in place

Highlight anything, tell Cephable how to change it, and it rewrites, shortens, or translates — right inside your app.

Automate

Agentic workflows

Describe a multi-step job in one sentence. Cephable navigates your apps, drafts, updates records, and reports back.

Everywhere

Works across every app

Cephable isn't locked to one program. It works system-wide — your browser, your editor, your inbox, all of it.

Universal

Universal by design

Voice, head tracking, facial expressions, switches and more — more than one way to drive your computer, built in from the start.

Private

On-device & private

Every model runs locally. No cloud calls, no data collection, works offline. Your words and work stay on your machine.

Frequently asked questions

Questions about Cephable

About Cephable

Cephable is a private AI assistant for Windows and Mac that runs on your own computer. It controls your apps by voice or command, and it writes and rewrites text in any app. It also completes multi-step tasks, like researching a topic, building a Word document, and drafting the follow-up email. All of the AI processing happens on your device, not in the cloud.

On-device AI runs the model on your own computer instead of in a vendor's data center. Cloud assistants like ChatGPT and Copilot send your prompt and its context to remote servers and charge by usage. Cephable runs a fine-tuned small language model locally. Your work stays on your machine, keeps running offline, and never adds per-token fees.

For most daily work, yes. Cephable handles drafting, summarizing, file work, web research, and multi-step tasks across every app on your computer, not just one vendor's ecosystem. Frontier cloud models are still faster and stronger at heavy reasoning and image generation. Many teams make Cephable their default and keep a cloud model for the exceptions.

Privacy and cost

Yes. Once you sign in and your model downloads, Cephable's voice controls, writing tools, and automations all run without an internet connection. Steps that need the web by nature, like online research or sending an email, need a connection for that step only. Cephable never needs a cloud AI model to do its work.

No. Cephable processes your prompts, files, and results on your device. Your task history is encrypted to your Cephable account and your computer's user profile, and it isn't stored on Cephable's servers. Your data never reaches a cloud model, and Cephable never uses your data or product telemetry to train models. Cephable is private by architecture, not by contract.

Cephable has three plans. Basic is free and includes Control. Plus adds Create, and Pro adds Automate. No plan charges token or usage fees, and teams get volume per-seat pricing.

Getting started

No AI PC is required. Cephable runs on Windows and Mac computers with Intel, AMD, Qualcomm Snapdragon, or Apple silicon processors. Control and Create run on 8 GB of RAM, and Automate needs at least 16 GB. You don't need a dedicated graphics card. Newer hardware simply runs tasks faster.

Download Cephable from cephable.com or the Microsoft Store and sign in. During setup you choose the kinds of work you do, and Cephable suggests prompts that fit that work. It also detects your hardware and installs the model optimized for it. The Explore tab has step-by-step guides and videos for common tasks.

Last updated