Local-first pocket AI recorder

The AI pocket recorder that never sells your voice.

Built on the Waveshare ESP32-S3 with a 1.8" AMOLED touch display. Raw audio lands on your MicroSD first, transcribes on your localhost Whisper server, and syncs directly to Obsidian and MemoryHive.

What it does today

Designed for private capture and instant recall

Meeting & Voice Recorder →

One-press recording to MicroSD with local Whisper transcription on your localhost, structured summaries, and action items synced to Obsidian & MemoryHive.

AI Agent Voice Assistant →

Hold-to-talk Q&A grounded in your notes and worklogs. Runs on your localhost by default, with explicit per-session approval before any cloud model is used.

31 Watch Faces, Music & Tasks →

1.8" AMOLED touch UI with 31 custom watch faces, local MP3/WAV player, on-device tasks, Squatchy RF scanner, and MeshCore companion messaging.

Hardware blueprint

Waveshare ESP32-S3-Touch-AMOLED-1.8

MCUESP32-S3R8 240 MHz
Display1.8" 368×448 AMOLED
Memory16MB Flash · 8MB PSRAM
AudioES8311 + Digital Mic
StorageMicroSD (SDMMC)
Sensors / PMICQMI8658 · AXP2101
Plans & Hardware Kit (AGPL-3.0)

Own your data, or let us host your sealed container

Home (BYOD)

$29 / year
+ shipping and handling

Run the Pocket-AI server on your localhost (Mac, Linux, Pi, or NAS) with MemoryHive included.

Hardware Kit

~$99 kit + 1 yr
+ shipping and handling

Pre-configured Waveshare ESP32-S3 AMOLED kit including your first year of Hosted service.

Join the Pocket-AI Waitlist

Get notified when kits and hosted accounts open, plus firmware release updates.