Adoption
46%
of US adults own a wearable; 57% including connected devices.
Rock Health, survey fielded Dec 2025Otolo / 01 Problem
Body data got more useful just as it got harder to reach.
AI can now help people and teams make sense of what their devices measure. In 2026, the largest device makers made that data harder to reach.
1.1 / What changed
Three shifts, all in the last two years.
Access
2026
Oura and Strava restrict passing their data to AI tools, Garmin paused new developer access, and the Fitbit Web API shuts off on 2 November.
Google Health developersAI
MCP
The Model Context Protocol gives AI assistants one open way to use outside data and tools. Claude, ChatGPT and local models support it.
Model Context Protocol1.2 / The problem
What people and teams run into today.
P.01
Locked in the vendor's cloud
Readings go to the device maker's servers before they come back to you. What you can do with them depends on that company's API, its terms and its roadmap.
When an API is retired or its terms change, studies and apps built on it stop working.
P.02
A score, not the signal
Most apps return a summary score. The raw measurements and the method behind the score are hard or impossible to get.
Teams cannot compare devices, and people cannot check what a number actually measures.
P.03
No history you can move
Switching devices usually means starting over. Using the data with your own AI means routing it through yet another company's cloud.
There is no common, local place where readings from different devices live together.
1.3 / Who has it
Three groups feel it first.
We start with teams that already pay for body data and feel vendor lock-in most sharply.
FirstPaid
Research labs and brand R&D teams
Labs running studies, and brand teams that test products on people.
- They need
- Consistent raw data across many people and devices, on a pipeline no vendor can switch off mid-study.
- Today
- Each vendor's cloud, export format and terms, stitched together study by study.
Also firstPaid
Enterprise teams learning AI agents
Companies training staff to build working AI agents.
- They need
- Practice data real enough to matter and safe enough for security and HR to approve.
- Today
- Courses on sample data that rarely turn into working agents.
Open coreFree
Developers and researchers
Builders and labs who want raw, portable data and their own choice of model.
- They need
- Readings in open formats on their own machines, and an AI layer they control.
- Today
- One vendor or one operating system at a time, through fragile workarounds.
1.4 / Where we are
Built from customer discovery.
Otolo is the result of more than a year of customer discovery.
Earlier concepts were tested in hundreds of customer interviews and refined into the current platform. We are in active conversations with brand R&D teams about how they test products on people, and a working prototype runs today.