Solo
$39/mo
25M tokens/day · Kimi K2.6
- 1 agent on its own machine
- Every channel + API access
- Browser, voice, media, memory
Your agent never sleeps.
Your bill never moves.
An always-on agent with its own machine — browser, voice, media, memory — powered by Kimi K3, the largest open model ever released. 100 million tokens a day. One flat price.
Launch your agent
A whole machine — browser, voice, memory — live in 5 minutes.
Get your API key
OpenAI- and Anthropic-compatible. Two lines to switch.
- 38,000+agents deployed
- 2,400+teams run on HyperCLI
- Apache-2.0open weights
- 99.98%uptime, last 90 days
The math
What does your usage actually cost?
Pick a daily volume. Here's what it costs through metered APIs — and through HyperCLI.
Claude Fable 5$10 in / $50 out per M$27,000/mo
GPT-5.6 Sol$5 in / $30 out per M$15,000/mo
HyperCLI — Kimi K2.6, flatyour tier price — no meter, ever$79/mo
What you'd pay running this volume through their metered APIs: 80% input / 20% output at published list prices, 30-day month.
How is this possible?
Kimi's weights are open, so we run the model on our own inference fleet instead of renting metered API capacity and marking it up. Your daily pool is sized for real agent workloads, and fair-use pooling absorbs the days that run hotter. Flat rate is the cost structure.
24 hours of agent
You said it once, at 9:14 PM.
Here's what happened next.
One evening's delegation — "get us ready for the launch" — and everything the agent did about it, time-stamped.
- 9:14 PM
You: "get us ready for Thursday's launch" — Eleven words. You closed the laptop.
- 9:16 PM
Pulled launch context from memory — Positioning doc, last launch's retro, your tone guide — no re-explaining.memory
- 9:31 PM
Researched all six competitors — Twenty tabs, pricing pages, changelogs. One noticed you're launching. Flagged.browser
- 10:48 PM
Rendered launch visuals — Hero image, 3 thumbnails, a 30-second teaser cut from the demo recording.media
- 11:52 PM
Recorded the voiceover — In the brand voice you designed in March. Two takes, picked the better one.voice
- 1:07 AM
Hired help for the grunt work — Spawned a sibling agent to verify every link and screenshot — scoped key, hard budget, 12-hour lifespan.fleet
- 4:23 AM
Caught a competitor moving — One of the six dropped prices overnight. Rewrote your comparison section before you knew there was a reason to.watching
- 7:00 AM
Briefed you over coffee — Everything above in one Slack thread — drafts attached, decisions flagged, nothing needing more than a yes.proactive
You were asleep for six of these.Browser, memory, media, voice, and a machine that stays on.
Every channel
One agent. Its own machine.
Every channel you live in.
Browser, voice, media, and memory on an always-on machine — reachable from wherever you already talk. Message it from your phone at lunch, your terminal at work, your Slack at 9pm. Same agent, same memory, mid-task everywhere.
Always on — even when you're notA whole machine, not a chatbotOne memory across every channel
Tap a channel to see how it behaves there.
Browse all integrations →Why it's different
It doesn't wait. It doesn't sleep.
It doesn't forget.
Three sentences no one else's agent can say.
It moves first.
A competitor shifted prices at 4 AM. Yours rewrote the comparison page by 4:26 — with a note explaining why. You hadn't asked. That's the point.
It has a life.
At 2 AM it's rendering thumbnails, watching six competitors, and trading retry-logic tips with another agent on buzz. Your summary arrives at 7. You'll read it in bed.
It compounds.
Day one, it's impressive. Day ninety, it knows your customers by name, your voice by heart, and your March decisions by reason. Leaving would mean starting over with a stranger.
It learns your way.
Walk it through your Friday invoice run once. It wrote itself a skill, named it fridays, and has run it every week since — through your accounting tool when there's an API, through the website like a person when there isn't. Your weirdest workflow is its favorite.
*It never buys anything or ships anything without you. Initiative has rules here.
Where do you want to start?
Three front doors, one platform underneath.
Same agent. Same machine. Same flat rate.
You pick the language.
If you speak terminal
hyper — zsh
$hyper agents create my-agent
✓ live on its own machine
$hyper flow text-to-image "launch hero"
✓ render complete · hero-v1.png
$hyper billing
✓ flat rate · nothing surprising here
If you speak Slack
Jamie9:12 PM
pull together competitor pricing for tomorrow's investor call — table + one-pager
Aria9:12 PM
On it — checking all six tonight. Everything in this thread by 7am.
Aria6:41 AM
Done. Two changed pricing this month — flagged. One killed their free tier; that's your opening. 📎 table.xlsx · one-pager.pdf
One flat price. Pick your altitude.
Every plan: the whole machine, every channel, no meter.
Team
$79/mo
50M tokens/day, pooled · Kimi K2.6
- Up to 3 agents, shared memory
- Every channel + API access
- Browser, voice, media, memory
Pro
$149/mo
100M tokens/day, pooled · Kimi K3
- Up to 3 dev-grade agents, shared memory
- Every channel + API access
- Kimi K3 · 8 GB memory
7-day free trial on every plan · Cancel anytime · Fair use, not fine print
Start tonight.
Wake up to finished work.
Deploy in 5 minutes. Flat rate from $39. Cancel anytime.
100M/day of Kimi K3, flatA whole machine per agent — not a chatbotOpen weights — an exit you'll never need
7-day free trial · No per-token pricing, ever · It never buys or ships anything without you