TARX is not another chat seat. TARX_OS is the always-on runtime: goals, approval, Evidence. free software. Get a T-4 when the desk needs dedicated NVIDIA.
Founders, operators, and 2–15 person teams who already work on Macs and are done renting another chat seat.
TARX_OS on the Mac you already own. Conversation is the interface. The runtime stays on.
TARX software
Free
Unlimited local use on Apple silicon you have.
Typical 5-person chat mix
$280/mo
80% $20 seats, 20% $200, whole seats.
T-4 on the desk
$183/mo
$6,600 list ÷ 36.
What the category sells
A chat window
Most “private AI” for people is still a $20–$200 seat.
THE CATEGORY
What they sell. What we ship.
The market
The market sells private AI as a chat window with a privacy toggle. Literacy is prompting a rented model. The loop lives on someone else’s computer.
TARX
TARX_OS is the always-on runtime on silicon you own. Goals, approval, Evidence. free software. Get a T-4 when the desk needs dedicated NVIDIA.
Who this page is for
Founder / operatorHead of operationsSmall-team ITAnyone whose laptop already holds the work
What this buyer is already hiring for
Local LLM / on-device inference
Private Copilot alternative
AI ops without sending files to a vendor
Small-team automation that has to stay on the laptop
HOW TARX WORKS
How TARX works for people and small teams
01
Join on the Mac you have
TARX software is free. First-run stays local. Nothing attaches until you opt in.
02
Run one real goal
Conversation is how you talk to it. The runtime stays on — Spaces, Control, Evidence.
03
Approve the write
Consequential actions wait on Deny or Allow once. Read Evidence after. That is literacy.
04
Add silicon if the desk needs it
Get a T-4 for indoor or vehicle work when shared work outgrows the laptop. Same OS.
Start on the Mac each person already owns. TARX software is free and unlimited. First-run stays local. Conversation is how you talk to TARX. TARX stays on: Spaces, Control, Evidence.
Most teams only know chat seats. Put a real goal through Deny or Allow once. Read Evidence after. That is the product — not another subscription window.
When shared work needs dedicated NVIDIA, get a T-4 for indoor or vehicle work, indoor / vehicle. Same OS. Same approval. Join the Supercomputer. Or own Private Hardware.
THE WORK
WHAT PEOPLE ACTUALLY RUNS.
Daily work that cannot leave the laptop
Drafts, files, and context stay on this Mac. LOCAL by default.
A team that already has Macs
Same OS on every seat. No per-person license. Supercomputer only when shared work needs more.
Approval before a write
Consequential actions wait on Deny or Allow once. Evidence after.
ECONOMICS
Five people. Public seat math versus silicon you own.
Path
Five people
What you actually get
Typical chat mix
$280/mo
Rented windows. 4 × $20 + 1 × $200.
All-Pro floor
$100/mo
Still a seat. Still a hop.
TARX_OS
$0
Always-on runtime on Macs you already own.
T-4 Supercomputer
$183/mo
Dedicated NVIDIA on the desk. List ÷ 36.
Seat prices are public. TARX_OS is included. Hardware monthly is list ÷ 36. Join the Supercomputer, or get your own.
THE STACK
Join the Supercomputer. Or own Private Hardware.
TARX_OS on the Mac each person already owns. Seats are $0. T-4 for indoor or vehicle work when the desk needs dedicated NVIDIA.
Placeholder
TARX_OS · Conversation into the runtime · free software
No. TARX software is free and unlimited on Macs you already own. There is no consumer seat.
What does a typical team spend on Claude instead?
Typical mix is 80% $20 seats and 20% $200 seats, counted as whole seats. Five people: 4 × $20 + 1 × $200 = $280/mo. All-Pro floor is 5 × $20 = $100/mo. A T-4 is $6,600 ÷ 36 = $183/mo.
When do we need a T-4?
When the team outgrows the laptop and needs dedicated NVIDIA capacity on the desk.
TWO EQUAL PATHS
JOIN THE SUPERCOMPUTER. OR OWN PRIVATE HARDWARE.
Join the Supercomputer. Or own Private Hardware.
JOIN · TARX_OS
Join the Supercomputer
Give TARX superpowers.
TARX works privately on your computer. Connect to the Supercomputer when you want TARX to research the web, use more skills, and take on bigger jobs with deeper models. Supercomputer connection is in beta, and capacity is being qualified.
