Your own Discord agent.
Runs the second you plug it in.
highermark is a tiny ESP32-S3 device that hosts Curtana, your AI agent, and Bard, a real Discord music bot. Plug it into any machine and both come online — joining your voice channel, playing music, recording and organizing data right on the device.
> pay once · own the code · name your bot · connect any LLM · take it anywhere
password
Curtana
the AI agentCurtana is your agent layer. It doesn't run a model on the chip — it connects to one: bring your own key for any LLM (OpenAI, Anthropic, Mistral, Groq, local). Wire it to your OpenClaw or Hermes agent flows and let it act in your server.
> tested with OpenClaw joining a live Discord channel.
Bard
the music botBard joins your voice channel and plays from YouTube, Spotify, Twitch and more. Queue, skip, loop — and adjust the volume freely, because nothing here is paywalled. Two years of flawless uptime; now it ships on a device you own.
A whole bot host, the size of a watch face.
Plug in → it's live
USB-C into your laptop, a Raspberry Pi, a home server — anything. The board boots, Curtana and Bard start, and the bot joins your server. Unplug and it stops cleanly. The 1.69" IPS screen shows status, now-playing and the active channel.
Pair your Discord token + LLM key in the dashboard.
Both bots auto-start on power. No services to babysit.
Play music, run agent commands, record to the device.
Curtana connects to the models you choose
The chip is the agent runtime, not the model. Drop in an API key for your favorite LLM and Curtana routes prompts, tools and Discord actions through it. Hook it into OpenClaw or Hermes for richer agent workflows. Local, on-device inference is on the roadmap.
Bard is a no-nonsense music bot
Join a voice channel; queue from YouTube, Spotify or Twitch; skip, loop, and set the volume to whatever you want — no premium tier gating basic controls. Two years of uptime behind it.
Your data stays on the device
Recordings, logs and organizational notes are stored on the board itself as a stepping stone — a private scratchpad for your server that travels with the hardware instead of living in someone else's cloud. Want more room? Offload to Umbrel (see below).
Run it anywhere — or fork it
Keep it on the device, or run the same code on any Linux machine, computer or server you control. The source is yours to fork and modify. Plug into OpenClaw / Hermes and any LLM API as you grow it.
Name it. Skin it. Share it.
Call the bot whatever you like and give it any avatar/logo. Share it with family and friends across your servers — or keep it as your private AI assistant on the go.
Pay once. Own everything.
Most Discord bots charge every month — and gate basics like adjusting the volume behind a premium plan. highermark is the opposite: pay once, get the device and the bot, and own it all.
| highermark | typical bot service | |
|---|---|---|
| Pricing | one-time | $/month, forever |
| Adjust volume & controls | free, always | often premium-only |
| You get hardware | yes — a device | no |
| Sources | youtube · spotify · twitch | limited |
| Own / fork the code | yes | no |
| Self-host (Pi / server) | yes | rarely |
| Data location | on your device | their cloud |
| Bring your own LLM | any API key | locked / none |
Built on a proven ESP32-S3 board.
1.69" IPS LCD · 240×280 · 65K color · 32-bit dual-core LX7 · USB-C. ~41 × 33 mm. A pocket computer happy running headless as your bot host.
> reference hardware shown. founder pre-orders receive a prototype unit.
Bring your agents. Bring your storage.
Hand actions to OpenClaw and let it drive your server. Verified joining a live channel.
Route Curtana through your Hermes agent flows for richer automations.
Unaffiliated with Umbrel — but if you run one, it pairs nicely. Offload your AI history to Umbrel's extra storage at your convenience (in testing), plug into the agents you host there — OpenClaw, Hermes — and use it for file storage.
[ works great with Umbrel ↗ ]Bard, battle-tested
Two years of flawless music-bot uptime. OpenClaw verified joining a live channel.
Curtana on the device
Plug-and-run agent + music bot, dashboard, BYO-LLM keys, on-device data.
OpenClaw / Hermes & multi-LLM
First-class agent integrations, Umbrel offload, one-click model switching.
Local on-device AI
Small local models running on the chip — no cloud needed for basics.
Companion products
Round-LCD variant, adapters, and more in the founder bundle.
Bard has run two years without missing a beat. Now Curtana joins it on a chip you own — plug it in and your Discord agent is live. Pre-orders + roadmap below. ▸◆▌
↳ paste your real tweet URL into index.html and uncomment the widgets.js script to render the live embed here.
[ contact us on X ↗ ]Pre-order for $100.
Your $100 supports the build and unlocks full functionality. You get one prototype device, your spot on the v1 waitlist, the newsletter, and a suite of companion products as they ship.
- ✓ one prototype device
- ✓ v1 waitlist + newsletter
- ✓ full functionality to test
- ✓ Curtana + Bard, name them anything
- ✓ suite of companion products
Other ways to support: $2 X Multiplayer access · $25 Curtana early-supporter tier — see X Multiplayer page or project breakdown for all contribution tiers.
