Skip to main content
Back

Open-Source Omi Hits 13k Stars Building an Always-On AI Memory Wearable

Omi bundles a wearable, mobile apps, and a self-hostable Python backend into a single MIT-licensed stack for always-on transcription and AI recall.

Repo, Full Access
  • Omi is an MIT-licensed AI companion covering wearable, mobile, desktop, and backend layers.
  • One-command --yolo setup launches the macOS app against a hosted backend with zero configuration.
  • Backend uses FastAPI, Deepgram, Pinecone, Redis, Silero VAD, and speaker diarization on GPU.
  • Reports LoCoMo 86.6% and LongMemEval 83.3% on long-term conversational memory benchmarks.
  • Companion $89 pendant and ESP32-S3 Glass dev kit with open hardware designs.
  • Includes MCP server, REST API, and SDKs in Python, Swift, React Native, plus 250+ community apps.

An always-listening AI companion sounds like a proprietary product pitch, but Omi is shipping the whole thing as an open-source stack. The project has crossed 13k GitHub stars and now bundles a Flutter mobile app, native desktop clients, a Python backend, firmware for a pendant and smart glasses, and multi-language SDKs, all under an MIT license.

Omi captures your screen and conversations, transcribes in real time, generates summaries and action items, and offers an AI chat that remembers everything you've seen and heard. The pitch is a second brain that indexes everything you say and see, then answers questions about it later. The maintainers claim 300,000+ users across desktop, phone, and wearables.

One command, no credentials

The friction for self-hosting has dropped sharply. A single clone-and-run command spins up the macOS client against the cloud backend:

git clone https://github.com/BasedHardware/omi.git && cd omi/desktop/macos && ./run.sh --yolo

That builds the macOS app, connects to the hosted backend, and launches. No env files, no credentials, no local backend. Windows users get a similar path with npm run dev against a public config, and the mobile app builds via