Your voice stays home
Speech recognition and voice synthesis run on the box itself. Your microphone stream turns into text right there and goes no further.
Give your agent a voice.
Your voice never leaves the box.
Teaport is a small, always-on computer for your OpenClaw agent, with ears and a voice built in. Talk to it and it answers in real time. Run the language model wherever you like, in the cloud or on your own hardware. Only text ever leaves the device, never your voice.
We reply with a payment link. No card needed to preorder.
Teaspoon AI is a member of NVIDIA Inception.
Ready out of the box
Plug it in. Join your Wi-Fi. Start talking.
Teaport arrives with your agent and its voice already installed. No terminal, no flashing, no lost weekend.
- 01
Plug it in
Give it power. It boots ready to talk, and it stays on.
- 02
Join your Wi-Fi
Connect it from your phone. No keyboard, no monitor.
- 03
Add your LLM and talk
Paste in an API key, or point it at a model on your own hardware. Then say hello.
How it works
Everything but the LLM runs on hardware you own.
Most “private” voice assistants still send your raw audio to the cloud. The rest cram the language model onto the same small device, and the agent gets dumber for it. Teaport runs speech recognition, voice synthesis, the agent and its memory on the appliance. The LLM runs wherever you choose.
On board isolated · low-power
- Teaport Engine · STT and TTS
- Teaport Brain · the orchestrator
- OpenClaw agent · persona · memory
audio + agent + memory live here
The LLM your call · swappable
- Any OpenAI-compatible URL
- A frontier cloud API → newest models
- Your own rig on the LAN
- → audio and text stay home
only the LLM call leaves the box
Your voice stays on board. Only text crosses the wire.
Speech to text and text to speech always run on the appliance, so your raw audio never leaves it. Where the text goes next depends on where you point the LLM.
With a cloud provider, only the text of what you say reaches them. Pick one with zero data retention. Run your models on your own hardware, turn off web tools, and nothing leaves your network at all, voice or text. Cloud-first assistants can’t offer that. Your data is their business.
What owning the box gets you.
The new cloud agents get a computer of their own, in someone else’s data center. Teaport gives yours one on your shelf.
Always on
Your agent gets a computer of its own, so it is there when your laptop is closed. It boots ready to talk and comes back by itself after a power cut.
Full duplex
Talk over it and it stops to listen. Teaport supports the OpenClaw realtime voice provider protocol, so it works with the OpenClaw browser and iPhone app clients.
Bring your own LLM
Use any LLM you like, local or in the cloud. Run it on your own hardware and everything stays home. Or pick a cloud provider with zero data retention.
Free up your LLM rig
An always-listening voice loop eats the memory bandwidth your LLM needs to decode. Move the loop onto Teaport and the model gets that bandwidth and RAM back.
Secure link, optional sandbox
The link between the plugin and the brain requires a secret token unique to your box. That won’t stop a prompt-injected agent from misusing the shell, file and web access it already has. For real containment, run it in a NemoClaw sandbox.
Talking to your agent
Your OpenClaw agent already handles the inbox, the calendar and the errands. Teaport lets you ask it out loud: from across the room, from a desk phone, or in a Discord channel. It answers in your language, with the personality you gave it.
A smart speaker of your own
Add the optional ReSpeaker XVF3800 microphone array and talk to Teaport from across the room, like a smart speaker. No phone, no app. Choose your own wake words. Until it hears one, nothing it hears goes anywhere. Microphone array and speaker not included.
Give it an extension
Connect it to your PBX or SIP trunk and it registers as an extension of its own. Dial it from a desk phone the way you’d dial a colleague, or route your existing call flows to it. Calls reach it through your phone provider like any call, and the listening still happens on Teaport.
You write its personality
You set your OpenClaw or NemoClaw agent’s personality in these files. Teaport gets them from your agent at the start of every session, so edit a line and the change is live the next time you talk.
Your own Discord voice bot
Tell it your server and who to follow. It joins your voice channel when you do and leaves when you leave. It’s your bot on your server, answered by your own Teaport. Strangers can’t add it or listen in. Discord carries the call audio, as with any Discord call, and the listening still happens on Teaport.
- English
- Mandarin
- Hindi
- Spanish
- Arabic
- French
- Portuguese
- Russian
- German
- Japanese
- Korean
- Italian
- Dutch
Thirteen languages, no picker
Talk in any of them and switch whenever you like. Teaport transcribes as you speak and answers aloud in eight of them.
Get your own Teaport
The Teaport appliance is $699 for the whole package: a Jetson Orin Nano 8 GB in a PANINI case by ROBOCO, with OpenClaw and Teaport pre-installed. Preorder now and we’ll reply with a payment link.
Rather build it yourself? Hobbyists, developers and enthusiasts can set Teaport up on their own hardware in five steps. Put it in a case of your own, or leave it bare on a shelf with the rest of your homelab.
-
Jetson Orin Nano (8 GB)
A Jetson Orin Nano (8 GB) flashed with JetPack 7.2 (L4T R38/R39, CUDA 13).
-
OpenClaw / NemoClaw agent
An OpenClaw or NemoClaw agent running on the device. The installer adds Teaport to it as a plugin.
-
A device to talk through
A phone, laptop, or any browser on the same network. Open the address the installer prints, accept the device’s certificate once, and start a Talk session. You talk through its mic and speaker. Nothing plugs into Teaport itself unless you add the optional microphone array.
-
A network and an hour
A home network and about an hour for setup and model downloads.
-
An LLM to point it at
Any OpenAI-compatible endpoint with tool calling. Use a cloud API key, or host a model yourself on your network (llama.cpp, vLLM, Ollama, LM Studio, NIM).
Building it yourself?
Your email gets you the install command now, and a note when new releases ship.
# one-line installer
$ bash <(curl -fsSL https://get.teaspoon.tech/teaport) We’ll email you when new releases ship. It’s free for personal use. Organizations need a written license.
License
Free for you. Licensed for organizations.
The engine is free for personal use by individuals under the Teaport Engine License 1.0. Organizations, and any commercial use of the engine, need a separate written license. The brain, built on Pipecat, ships under plain MIT. To license the engine, write to stefano@teaspoon.tech.