otolo

Otolo / 02 How it works

Capture locally. Store in open files. Serve to the AI you choose.

Five steps on hardware you already own. Otolo is the layer in between: device drivers, a local store and an MCP server.

2.1 / Architecture

Everything inside the dashed line runs on your devices.

Architecture and data flow. A supported device sends readings over Bluetooth to the phone, which syncs them to a local store of open files (SQLite and Parquet). The open-source Otolo MCP server reads that store on your computer and serves it to a local model, where data stays local, or to a cloud AI such as Claude or ChatGPT, which receives only what you select.
Readings are captured and stored on your devices; only what you select goes to a cloud AI.Open full size ↗
2.2 / Step by step

What happens to a reading.

  1. 01 CAPTURE

    Supported device

    HR · HRV · motion

    A wearable sensor with an open, documented interface streams readings over Bluetooth Low Energy. We start with devices whose makers publish their interfaces, and list every device we support.

  2. 02 RELAY

    Phone

    iOS · Android

    The Otolo app and its device drivers receive the readings on your phone and sync them to your computer.

  3. 03 STORE

    Local store

    SQLite · Parquet

    Readings land in open file formats on your computer. You can open them in any tool, back them up and move them, with or without Otolo.

  4. 04 SERVE

    MCP server

    Otolo Core · open source

    A local MCP server exposes your readings as tools an AI assistant can call, such as querying a metric for a date range. It runs on your computer.

  5. 05 ASK

    The AI you choose

    local model or cloud AI

    Use a local model, for example through Ollama, and nothing leaves your computer. Use Claude or ChatGPT, and only what you select is sent to that provider, when you ask.

2.3 / What stays, what leaves

Say exactly where the data goes.

A security reviewer should be able to read this table and find no gaps.

DataWhere it livesWho can see it
Raw readings and historyYour phone and computer, in open filesYou. Otolo does not receive them.
Questions to a local modelYour computerYou. Nothing is sent anywhere.
Questions to a cloud AIThe provider you choseOnly the results you select, under that provider's terms.
Otolo serversNot in the data pathNo one. Capture, storage and MCP run on your devices.
Employer programs (Otolo Cohort)Learners' own devicesProgram reports are aggregate and opt-in. No per-person body data.
We never sell user data or train models on it.
2.4 / Availability

What runs today, and what comes next.

CapabilityStatusWhat it means
Capture to local store to MCPWorking prototypeReadings flow from a supported device to open files and into the AI you choose
Open-source coreEarly 2027Free to install, inspect and extend
Supported-device listWith the open-source releaseEvery supported device listed, each with a tested driver
Accuracy dataPublished per deviceHow each device compares with a reference, with methods
2.5 / For developers

Plain files and a standard protocol.

Example commands.

  • Formats: SQLite for the live store, Parquet for exports. Open in pandas, DuckDB or a spreadsheet.
  • Protocol: MCP, so any MCP client can use your readings, including a model running on your own computer.
  • Drivers: one per supported device, open source, with the firmware version tested and published accuracy data.
  • License: open source; the license is named at the first release.
2.6 / Questions

Common questions.

Which devices are supported?

We start with wearables whose makers publish open, documented interfaces, and add devices one tested driver at a time. The supported-device list ships with the open-source release.

Does Otolo see my data?

No. Capture, storage and the MCP server run on your devices. Otolo does not receive your readings, and we never sell user data or train models on it.

Which AI can I use?

Any MCP client, including Claude, ChatGPT and local models run through tools such as Ollama.

What happens to my data if I stop using Otolo?

Your readings are in open files on your own computer and the core is open source, so your data and the software stay with you.

Can we run a multi-device study?

Yes. Otolo for Research brings readings from every participant's supported devices into one consistent dataset in open formats, with the device and firmware recorded for each reading.

What does an employer see from a cohort?

Only aggregate, opt-in program results. No per-person body data reaches the employer.

Is this a medical product?

No. Otolo is a data and developer platform. It moves and stores measurements; it does not interpret them for medical purposes or give medical advice.

We are working with a small number of teams on the first release.