Kiyomi cherry blossom circuit mark
Kiyomi · Private · US-based

Your own AI. One key. One line.

A frontier-class model served from hardware we own in Atlanta. Buy a key, paste one line, and it becomes your own AI in a desktop app. One flat price. Your data never leaves our machines.

endpoint https://kiyomibot.ai/v1
Runs on hardware we own Your data never leaves One flat price, no meter
~ your terminal
One flat price, no meter

Pricing that beats the $200 subscription.

Three plans, each licensed to one person. A flat monthly rate with a daily quota, not pennies-per-token that punish you for working.

Pro
$25/mo
Serious daily work, one person.
  • 2,000 requests / day
  • 60 requests / minute
  • Priority queue
  • One-line setup, key inside
Get Pro
The $25, not $200
Power
$75/mo
Full-time power users.
  • 10,000 requests / day
  • 200 requests / minute
  • Priority queue
  • One-line setup, key inside
Get Power
Still 62% under Claude
Your one-liner
The moment you buy, this appears with your key already inside:
KIYOMI_API_KEY=sk-••••••••••••
bash -c "$(curl -fsSL https://kiyomibot.ai/install.sh)"
Your purchase page has the free desktop app download and this line. Paste it in Terminal or PowerShell (or hand it to your AI) and the app opens on Jah 3.0. That is the whole setup.

Checkout takes under a minute Your one-liner + key appear instantly 30-day money-back guarantee Cancel anytime

Enterprise Teams, dedicated capacity, custom quotas, your own private instance. Talk to Richard →
The hardware

One rig in Atlanta. Owned outright.

This is not a reseller skin over someone else's cloud. Your requests are served by machines we bought, built, and run ourselves.

Serving live · Atlanta, GA
High-performance GPU with triple cooling fans
GPU compute
Server cabinets with red cabling
The racks
RTX PRO 6000
1M
Token context
276/s
Tokens streamed
100%
Owned, zero cloud

Why it matters: when the hardware is owned, the economics are flat and the privacy is physical. Nobody else's terms of service sit between you and your model.

The model inside

Jah 3.0. Open weights, no secrets.

Kiyomi runs on Jah 3.0, our own model, and the exact weights serving customers right now are public. Open, downloadable, verifiable. Don’t take our word for what powers your work: inspect it.

304B
Parameters
1M
Token context, verified
~276 /s
Tokens on our rig
MIT
License. Truly open
🤗 Hugging Face
RMDWLLC/Jah-3.0
167 GB download · 48 safetensors shards · FP8/FP4 · vLLM-ready
Download the weights →
Hardware to run it yourself: the weights need roughly 256 GB of fast memory as a floor. Two NVIDIA DGX Sparks linked (256 GB) run it. A 256 GB Mac Studio holds it with conversion. Full speed and the whole 1M context is a 4× RTX PRO 6000 class rig (384 GB), which is exactly what we serve on.

No rack at home? That’s the API: the same Jah 3.0, on our hardware, one line away.

Enterprise

Your team. Your quotas. Your own instance.

Dedicated capacity on our hardware, custom rate limits, and a private deployment shaped to your stack. Tell us what you need and Richard will reach out directly.

Goes straight to the founder. No sales team, no drip campaign.

Atlanta, Georgia, USA

Your tokens never leave the building.

Every request is answered from machines we own in Atlanta. No foreign clouds. No third-party model providers. No one training on your work.

33.7490° N, 84.3880° W · Serving now
After you buy

Four steps. Two of them are clicks.

Buy your key, download the free desktop app, paste your line. The app opens with Jah 3.0 already selected — nothing to pick, nothing to configure.

# Then paste your line. Mac & Linux — Terminal
KIYOMI_API_KEY=sk-your-key bash -c "$(curl -fsSL https://kiyomibot.ai/install.sh)"
 
# Windows — PowerShell
$env:KIYOMI_API_KEY="sk-your-key"; iex (irm https://kiyomibot.ai/install.ps1)
 
# It sets your key, verifies it live, and opens the app on Jah 3.0.
+ Start with a goal Jah 3.0 · XHigh

Step 4: the app opens like this — Jah 3.0 already selected. Type what you want. That’s your private AI, working.

Or skip the setup. Hand it to your AI.

Copy this prompt, drop your one-liner in, and paste it into the AI you already use. It runs the install for you and confirms the app opened.

Set up my private Kiyomi AI on this machine. 1. If the Hermes desktop app is not installed, download and install it from https://hermes-agent.nousresearch.com (Mac or Windows button). 2. Then run my setup command (from my purchase page or email): [PASTE YOUR ONE-LINER HERE] in my terminal (macOS/Linux) or PowerShell (Windows). It configures the app with my key and verifies it live. Confirm the app opened with Jah 3.0 selected when you are done. Never send my key anywhere except https://kiyomibot.ai/v1.
Paste your one-liner in, then give the whole thing to Claude, Codex, or any assistant.
From zero to private

Three steps. No migration.

01

Get your key

Pick your plan: $25, $50, or $75, each for one person. Your key, your line, and the app download buttons appear the second you pay.

02

Run one line

Download the free desktop app (one click), then paste your line into Terminal or PowerShell — or hand both steps to your AI. The app opens on Jah 3.0.

03

Build in private

Your prompts, code, and data run on our Atlanta rig and nowhere else. Flat bill. No meter anxiety. Just work.

Why Kiyomi

Own your stack. Keep your data.

The independence of a local model, without the hardware. Served from machines we own, in the US.

US-based & private

Your prompts and code run on hardware we own in Atlanta. Never routed through foreign clouds, never sold, never trained on.

Zero lock-in

Under the hood it is a standard API key, OpenAI and Anthropic compatible. You are never trapped in anyone’s app, including ours.

Flat, not metered

One price a month with a daily quota. The bill never grows because you used it well. No surprise token charges.

Built by the owner

Made by an engineer who runs his own AI on his own hardware, not a reseller skin over someone else's cloud.

Serious capacity

Served from a 4× RTX PRO 6000 rig. Built for real agentic workloads, not toy demos.

Fair-use protected

Daily quotas keep the rig healthy for everyone. No unlimited plans that get throttled into uselessness.

Questions, answered straight

Asked before. Answered honestly.

Yes. Your requests run on hardware Kiyomi owns and controls in the US. We never sell your data, never train on it, and never route it through a third-party model provider. Connected features disclose when they contact another service.

Their APIs are metered, so the bill grows with every token, and your code runs in their cloud. Kiyomi is a flat monthly rate with a daily quota, served from our own US hardware. Your same workflow, your own model, no meter.

No. After you buy, you get two buttons and one line: download the free desktop app (a normal installer, like any other app), then paste your line into Terminal or PowerShell — or hand both steps to the AI you already use. The app opens with Jah 3.0 already selected. If you can download an app and copy-paste, you are set.

They keep the rig healthy for everyone and stop a single account from monopolizing capacity. One heavy day never kills your month. The per-minute limiter guards concurrency, the daily bucket gives you room to work.

Rare, but real: some antivirus setups block the desktop app’s installer. Your key still works either way — the terminal command hermes runs Jah directly, and the key plugs into any OpenAI or Anthropic compatible tool. And you can always email hello@rmdwllc.com: you are talking to the founder, and I will personally get you running.

Every paid plan comes with a 30-day money-back guarantee. If it isn't for you, email us within 30 days and we refund in full. No questions, no hoops.

Stop renting intelligence. Start owning it.

Get your key, paste one line, and run your work on a model that's yours, for less than a tenth of what you'd pay Big Tech.

Get your API key
Looking for the Kiyomi web app? Your chats and builds now live at app.kiyomibot.ai