omi

What Omi records

Omi is an open source tool the README calls a second brain you trust more than your first. It captures your screen and your conversations, transcribes them in real time, generates summaries and action items, and gives you an AI chat that remembers everything you have seen and heard. The project states it works on desktop, phone, and wearables, so the capture is not limited to one device. The README also says Omi is trusted by more than 300,000 professionals, a claim the project makes about its user base. The point of the design is continuous context: rather than you manually logging meetings or notes, Omi builds a searchable memory of activity and then lets you ask questions against it. The chat side means the memory is interactive, not just a log you scroll. For people who take a lot of meetings or switch between many apps, the value is having a single place that saw the screen and heard the call, then can summarize what happened and what to do next without the user reconstructing it from scattered sources. The real time transcription is what makes the memory useful the moment a meeting ends, because the summary and action items are already drafted from the captured audio and screen rather than waiting on a later manual write up by the user.

Hardware and software together

Omi is unusual in pairing open source software with open source hardware. The README describes wearables that pair with the mobile app for more than 24 hours of continuous capture, and it links to an Omi device and an Omi Glass dev kit built on an ESP32-S3 with camera and audio. The hardware designs are published as open source, with a buying guide and assembly docs. On the software side, the repository holds desktop apps for macOS and Windows, mobile app setup, and a backend. The macOS quick start clones the repo and runs a script with a yolo flag that builds the app, connects to the cloud backend, and launches it with no env files or credentials. Windows uses npm with an example env file. A make setup command installs Git hooks and syncs a pinned backend Python environment for pre push checks. This blend of a wearable capture device and a cross platform app is what lets Omi record conversations beyond the laptop, which a pure software tool cannot do as easily. The ESP32-S3 based dev kit shows the hardware is approachable for makers, and the open designs mean a user can repair or modify the device rather than rely on a closed consumer product that ships as a black box.

Building and extending

The README points developers to a set of guides and SDKs. There is an app development guide with example apps such as GitHub, Slack, and OmiMentor, plus audio streaming apps and custom chat tools. An API reference covers REST endpoints for memories, conversations, and action items. Device SDKs are published for multiple languages, with shared BLE UUIDs and packet framing for TypeScript, Go, Rust, C++, and Dart, and fuller BLE plus Opus plus Deepgram support in the Python device SDK, plus Swift and React Native SDKs. An MCP Server adds Model Context Protocol integration so the memory can be wired into other AI tools. The backend deep dive, transcription pipeline, chat system, and BLE protocol are each documented for engineers who want to understand or modify the internals. The project is written in Python and is released under the MIT license. For a builder, the combination of documented APIs, multiple language SDKs, and open hardware means Omi can be extended as a platform rather than used only as a finished app. The MCP server in particular is notable because it lets the captured memory become input to other agents, so Omi acts as a source of context for a wider toolchain instead of a closed notebook that nothing else can read.

Editorial conclusion

The project is released under the MIT license, written in Python, and hosted at github.com/BasedHardware/Omi.