Big Tech AI listens to everything. ChatGPT logs your conversations.
Claude reads what you ask. Gemini knows what you wonder about at 2am.
All of it sent to servers you'll never see, owned by companies that
sell your data.
JeffyAI doesn't.
This isn't a corporate product. JeffyAI was designed by one person —
Jeff Johnson — in the Pacific Northwest, and built with the help of
a few cooperative AI partners. We did it because no major AI product
would guarantee what should be obvious: your conversations
are yours. Take that, Silicon Valley.
"I built JeffyAI because I wanted an AI that worked for me — not for
the company that made it. That meant local-only, no cloud, no
surveillance. So I built it that way. Apparently nobody else would."
— Jeff Johnson, founder, Sagentica
How JeffyAI actually works
Where it runs
On your computer. Not in a data center. Not in the cloud.
Not on Sagentica's servers. The AI brain (a local language model) plus
JeffyAI's privacy infrastructure (memory, vault, profiles, license signing)
all run on your local hardware. You can install it on a desktop with no
internet connection.
What we see
Your email address (for license delivery). Your purchase
amount (for tax records). Optionally — only if you opt in — minimal crash
reports or anonymous usage stats. We literally cannot see your conversations
because they stay on your machine. There is no cloud sync, no
backup-to-our-servers, nothing. Full privacy policy →
The AI engine
JeffyAI uses open-source AI models (like Qwen, Llama, Mistral) running
locally on your hardware via Ollama.
But the AI model is just one part — the privacy infrastructure around
it (memory, vault, multi-profile, license signing, agentic mod system,
edition-aware features) is what makes JeffyAI a product. We call that
infrastructure QAICX (K6). More on that below.
What you pay
$99 founders / $149 list, once. Yours forever. The version you buy
is yours to use as you like — even five years from now. First year of updates
included. After that, want new features, new versions, major upgrades? An optional
small annual fee. Skip it and keep using what you bought — it doesn't expire.
Critical security fixes are free during your version's supported lifecycle
(AI moves faster than traditional software — support windows for AI products are typically
shorter than 10-year Office runs). If you're online, patches arrive automatically. Off-grid
users pay shipping for physical delivery. (We're not renting a helicopter to find Prepper Joe.)
The local-first design means even an unsupported version keeps running on your machine —
your data and memories transfer to whatever you upgrade to. If we change pricing in the future,
the change applies to new purchases — not yours. That's a promise written into the License Agreement.
What you can do
Chat with Jeffy (the AI character). Create AI personalities for your
household. Encrypt private notes in a local vault. Add voice
interaction. Customize the personality. Activate add-ons
(sub-accounts, white-label, privacy controls suitable for medical
contexts — not certified). Everything happens on your machine.
Automatic Transmission — the brain swapper (v1 roadmap — in development)
What it does
JeffyAI runs not one AI model but several specialized ones, and routes
your query to whichever one is best at it. Code questions get answered
by a code-trained model. Creative writing goes to a chat-tuned model.
Factual lookups engage an anti-hallucination layer. Casual conversation
routes to a smaller, faster model for snappy reply time. The selection
happens in milliseconds. You never think about it.
Why it matters
Cloud AIs (ChatGPT, Claude, Gemini) use one giant model for everything.
That's expensive, slow, and a one-size-fits-all answer for queries that
aren't one-size-fits-all. Local consumer AIs that try the same trick
usually grind on small hardware. JeffyAI's Automatic Transmission stays
responsive by running the lightest specialized model that can answer
each query well — and only that one.
Hardware-aware
The system detects your RAM, CPU cores, and GPU on first run and tiers
itself accordingly. A 4GB Celeron laptop gets a lightweight 1.5B
configuration. A 16GB premium machine with integrated GPU gets the full
multi-model stack. A 32GB workstation with discrete GPU gets bumped up
a tier. The same software, intelligently configured to your machine.
Graceful fallback
If only one model is installed, Automatic Transmission falls back to
single-model behavior. Nothing breaks. As you add more specialized
models (one-time pulls from Ollama), the routing automatically expands
to use them. You don't learn anything new — the system gets smarter
on its own.
The brand stack, plainly
Sagentica
The parent company. Based in Oregon. Founded by Jeff Johnson. Owns the QAICX platform.
QAICX (K6)
"Quality AI Customer Experience" — Sagentica's proprietary multi-tenant private AI platform. The infrastructure inside JeffyAI. Pronounced kay-six.
JeffyAI
The consumer product. The first product built on QAICX. What you install on your computer.
Jeffy
The AI character inside JeffyAI. Who you actually talk to. ("Hey Jeffy, what should I make for dinner?")
Building a private AI product yourself? Run it on QAICX.
QAICX (K6) is Sagentica's white-label private AI platform. Healthcare
companies, legal firms, education platforms, therapy apps — anyone who
needs their own branded AI product without building the privacy + license
+ multi-tenant infrastructure from scratch — can license QAICX and ship
on top of it.
Your customers see your brand. Your data stays in your tenant. Sagentica
handles the platform.
Big Tech AI products charge $20-30/month for cloud-hosted assistants
that read everything you type. Enterprise "private AI" products charge
$500+/year and require an IT department to install.
JeffyAI charges $99 once (founders sale, list $149). Runs entirely on your
computer. Setup takes 5 minutes. Privacy is the default — not a
premium upsell. Agentic AI for everyone, at the price point of
a video game.
Knows you. Tells nobody. Local. Loyal. Lifelong. This is what
"democratization of AI" should have meant.
Why this matters
In 2024-2026, every major AI company has been caught in some form of
privacy scandal — chats leaked, conversations used to train, data sold
to advertisers, government requests honored. Every "we don't store your
data" claim has been quietly walked back when it became inconvenient.
JeffyAI is built on architectural commitment, not
policy promise. Your data can't be leaked because it never left your
machine to begin with. The privacy isn't a feature you trust us to
honor — it's a fact that's true regardless of what we say.
Who built this
Jeff Johnson. Solo founder. Pacific Northwest. Built JeffyAI alongside
the QAICX platform over many months of nights and weekends, on top of
a long career building things on the internet.
No VC. No engineering team. No marketing department. Just one person
who was tired of Big Tech AI listening to everything, and decided to
build the alternative.
If JeffyAI helps you, that's the whole point. If it doesn't, the
download is free. If you have questions, email
[email protected] — it goes
directly to me.