A memory-keeping assistant on your own hardware — voice-first, teaches the kids, and never sends your data anywhere.
Notes, decisions, documents — recalled when you need them.
A voice assistant that greets you and gets things done.
Maths, physics, history, languages — 1:1, never just the answer.
Local-first storage, anonymisation, tamper-evident logs.
Personal space per person — kept separate.
Seal & archive papers so they're verifiable for years.
The technical capabilities behind this — from the live feature inventory.
Episodic, semantic, procedural, reasoning and associative memory across 5 data layers; measured recovery — 100% of decisions recovered vs 43% without, 99.94% compression.
Chat, files + OCR, export to PDF/DOCX/XLSX/MD/HTML, skills, subagents, 31 languages, per-turn anonymisation; per-user language over one shared memory.
Define a prompt + scope and run it on a schedule or trigger (time, file change, memory event, webhook), optional auto-proceed, on the model profile you pick — including local Ollama. Replaces Zapier/cron/n8n glue on your own infra.
One router speaks OpenRouter (cloud), local Ollama and native SDKs with automatic fallback; your data stays on local storage and the cloud model only ever sees anonymised text.
A mic button transcribes speech locally (Whisper); audio is deleted immediately and never stored.
Free engine on your own hardware. No credit card. Founder pricing through 31 October 2026.
View pricing Back to home