Build your intelligence layer on wearable data
Open Wearables gives your team a ready AI agent for wearable health data. It runs as a service in your own deployment, works with the LLM provider you connect, and becomes the intelligence layer your product competes on once you add your voice, tools and data.
Built on Open Wearables
Overview
One agent module for teams building AI features on wearable data
Everything a product team needs between the data in Open Wearables and an AI feature in their app, delivered as one module that runs in your deployment. These are the parts you would otherwise build yourself.
- 01
Tools for wearable data.
Ready tools query sleep, activity, heart and workout data through the Open Wearables schema, for the user and time range a question needs, so the model does not get the whole history in its prompt.
- 02
A prompt profile you can edit.
The prompt is split into blocks. Change the tone, your platform's context and how the data is analysed, while the instructions for using the tools stay as they are.
- 03
Your own tools and data sources.
Register tools that read your labs, nutrition logs, training plans or product catalogue, and the agent uses them next to wearable data.
- 04
One core for every AI feature.
Chat, scheduled summaries, post-workout reviews and summaries for the screen in view are all requests to the same agent, so you build the logic once.
- 05
Evaluation built in.
A question set and judges score the agent's answers. Run it after you change the model, the prompt or a tool, and compare the runs side by side.
What you get
Why teams start here instead of from scratch
A working baseline from day one
The agent answers from the first request, and you tune it from there instead of assembling it from parts.
- Ready baseline
- Tunable
Any LLM provider
Connect the provider your team already uses. Switching models is a setting, and the evaluation shows whether the new one holds up.
- Model-agnostic
Runs in your deployment
A separate service beside Open Wearables, in your own cloud, reading the data where it already is.
- Self-hosted
- Beside Open Wearables
What you build on it is yours
Your prompts, tools and question sets are your product's logic, and they live in your deployment.
- Your deployment
How it works
From your deployment to an AI feature in your app
- 01
Run it beside your deployment
The agent reads the wearable data your Open Wearables deployment already holds. There is nothing new to collect.
- 02
Connect your model and set the scope
Choose the LLM provider, then decide what the agent talks about and how it sounds.
- 03
Add what only your product has
Register your own tools and data sources. They are the part of the agent that is specific to your app.
- 04
Measure, then ship
Test the answers on your own questions, then call the agent from your app with the screen or event in view.
Book a demo
What you can build
Insight that fits the moment
Each answer is shaped by where the user is and what just happened: the screen in view, the time of day, a workout that just synced. It ends with a step the user can take.
A morning summary with a plan for the day
Last night against that person's own recent nights, with a suggestion for today's training and a next question ready to tap.
Insight for the screen in view
Your app passes the screen and the agent stays on it: sleep on the sleep screen, training on the training screen, each with one thing to try.
A review when a workout syncs
The finished session against the user's recent ones, with what to change next time.
Answers that end in a next step
A question in chat gets an answer from the user's own readings and a step they can take, such as moving tonight's session.
Who is it for
Who builds on the AI agent
Digital health products
An AI feature that is yours, with its logic, voice and tools under your control.
Learn moreHealth data platforms
An agent you can offer the teams on your platform, running in the deployment you already operate.
Learn moreRemote patient monitoring
Summaries of each patient's trends for the care team, inside the scope you define.
Learn moreClinical and research builders
A setup you can test and measure question by question before it reaches anyone.
Learn moreCare providers
Guidance for members on data that stays in the deployment you promised it would.
Learn moreHospitals and telemedicine
A coach in your brand's voice, connected to your own training plans or product catalogue.
Learn more
One agent, whatever your users wear
Every source arrives in the same schema before the agent reads it, so an answer about last week's training works the same way whether the data came from a watch, a ring or a phone. You build one agent, and it answers about a person rather than a device.
Add the health agent to your product
The health agent is part of Pro, running inside your own deployment.
FAQ
Do you have more questions?
Is the AI agent available today?
It is a licensed module; request access and we set it up with you. If you want an assistant querying your data now, the MCP server ships today.
How do I build an AI health coach on wearable data?
Start from a deployment that holds your users' data, then put an agent in front of it that knows how to query that data. The AI agent gives you that agent ready, and you add your voice, scope and tools.
Can I use my own LLM?
Yes. The agent works with the LLM provider you connect, and changing the model is a setting.
Does our users' data leave our deployment?
The agent runs in your deployment and reads your data there. The only data that leaves is what goes to the model provider you connect.
How is this different from an agent framework like LangChain?
A framework gives you parts to assemble. The AI agent is assembled already, with tools for wearable data, a working baseline and an evaluation setup, and you configure it rather than build it.
How is this different from pasting our data into ChatGPT?
A general chat model sees whatever you paste and has to guess how to read it. The agent fetches the right slice of each user's data through tools built for it, inside the scope you set, and you can measure its answers.
How do I evaluate a health AI agent?
Write the questions your users ask with the answers you expect, run the agent on them, and score the answers. The agent ships with that setup, so you repeat it after every change of model, prompt or tool.
How is this different from the MCP server?
The MCP server lets your team's own assistant query your data. The AI agent is what you put in front of your users, inside your app.
Does it work with health scores?
Yes. It reads your custom health scores next to the raw signals.
Will it give medical advice?
It follows the scope you set. You decide what is out of bounds and test it on your own questions before anything reaches a user.
Can it connect to data that is not from wearables?
Yes, through tools you register. If you need tools for your own systems, such as labs, nutrition logs or EHRs, our team can build them with you.