KIVAGet notified
Launching on Kickstarter soon

A desk agent that actually pays attention

KIVA runs a GPT-class model on a small device beside your keyboard — wake it hands-free, talk to it continuously, and watch it answer with lifelike expressions and motion. Ask out loud. It does the work and looks up when it's done.

40% off reserved for the first 500·No spam — launch news only
0+
On the waitlist
$0
Early-bird price
Sep '26
Target ship
What KIVA does

Less a gadget, more a teammate parked on your desk.

Three things most smart speakers can't do — because KIVA was built around an agent, not a voice command list.

01

Sees what you see

Point it at your monitor and ask about whatever's on screen — a chart, an error, a contract clause. It reads context instead of waiting for keywords.

02

Always-on voice

A far-field mic array hears you across the room. Talk naturally, interrupt mid-sentence, follow up — no wake-word gymnastics, no phone in hand.

03

Gets the work done

Reminders, voice notes, drafting copy, answering questions, light desktop control. KIVA handles the everyday tasks out loud — then looks up when it's finished.

Why not just a smart speaker

A speaker answers. KIVA gets it done.

The difference isn't the voice — it's the agent behind it.

Smart speaker
KIVA
Interaction
Fixed command list + wake word
Open-ended agent — multi-turn, interruptible
Awareness
Hears one phrase at a time
Sees your screen + hears the room
Tools
A handful of built-in skills
Connects to your stack — mail, tickets, calendar, docs
Outcome
One canned action
Completes multi-step tasks, then reports back
Expression & motion

It reacts like something that's listening.

A round face that shifts with the conversation, and a head that turns toward your voice — the difference between a speaker and a presence on your desk.

Dynamic expressions

Dynamic expressions

KIVA's face listens, thinks, confirms, and reacts — lifelike micro-expressions that make every exchange feel natural and alive instead of transactional.

Expressive motion

Expressive motion

It turns to face you when you speak and nods along as it works. Subtle, lively movement that signals attention without ever getting in the way.

Make it yours

Your wake word. Your voice. Your KIVA.

Out of the box it's capable — but the personality is yours to set. Tune how it wakes, how it sounds, and how it talks back.

Custom wake word

Name it whatever you like. KIVA wakes hands-free to the word you choose — no fixed phrase, no button.

Response style

Concise or chatty, formal or playful — set how KIVA answers so it fits the way you actually like to be talked to.

A presence that's yours

Wake word, voice, and manner combine into a companion that feels personal — not a generic assistant everyone shares.

The hardware

Small enough to ignore. Capable enough to rely on.

KIVA from multiple anglesOn-desk scale reference
  • Display4" round LCD · expressive face
  • Compute6-TOPS NPU · on-device inference
  • ModelHybrid — local + cloud handoff
  • Audio4-mic far-field array · 3W speaker
  • Vision1080p wide cam · privacy shutter
  • Motion2-axis head · turns to face you
  • ConnectWi-Fi 6 · BLE · USB-C
  • Footprint92 × 92 × 140 mm
Backer tiers — preview

Reserve early-bird pricing now. Pledge when we launch.

Subscribers get first access to these tiers the moment the campaign goes live. Quantities are capped.

Tier 01 · Believer
$79
MSRP $149
−47%
  • 1× KIVA unit
  • Lifetime base agent features
  • Backer-only color: Graphite
  • Founders wall credit
Reserve access
Limited to first 200
Tier 02 · Early Bird ★
$99
MSRP $179
−45%
  • 1× KIVA unit
  • 1 year KIVA+ agent plan included
  • Choice of 3 colors
  • Priority Sep '26 shipping
  • Founders wall credit
Reserve access
Most popular · capped at 500
Tier 03 · Desk Duo
$179
MSRP $358
−50%
  • 2× KIVA units
  • 2 years KIVA+ for both
  • All colors unlocked
  • Priority shipping
  • Early beta firmware access
Reserve access
Limited to first 150
Where we are

A real product on a real schedule.

Done · Q4 '25

Working prototype

EVT hardware running on-device inference, voice loop, and screen vision end to end.

Now · Q2 '26

Pre-launch & waitlist

Building the community, locking tooling integrations, and finalizing the industrial design you see here.

Next · Q3 '26

Kickstarter campaign

30-day campaign. Subscribers get early-bird access first, then public tiers open.

Then · Sep '26

Production & ship

DVT → mass production. First units ship to early-bird backers.

Before you ask

The honest questions.

Is this just a smart speaker with a screen?+

No. Speakers run a fixed list of commands. KIVA runs an agent: it understands open-ended requests, sees your screen for context, connects to your tools, and completes multi-step tasks rather than firing one canned action.

What about privacy — there's a camera and mic?+

There's a physical privacy shutter over the lens and a hardware mute on the mic. Local inference handles most requests on-device; only what you explicitly send goes to the cloud. You control what each integration can see.

Do I have to pay a subscription?+

The base agent features work for the life of the device with no fee. The optional KIVA+ plan adds heavier cloud models and premium integrations — and early-bird tiers include it free for 1–2 years.

It's crowdfunding — what's the risk?+

Fair question. We already have working EVT hardware (not a render), a defined BOM, and a manufacturing partner lined up. We'll share teardown videos and factory updates throughout the campaign so you can judge progress for yourself.

Does subscribing commit me to buying?+

Not at all. The waitlist just guarantees you see the launch and early-bird pricing first. You decide whether to pledge once the campaign is live.

Be first in line when KIVA goes live.

3,200 people are already waiting. Early-bird pricing is capped — subscribers get the link before anyone else.

40% off for the first 500 backers