From Wearables to Wisdom: Building a Health Coach Agent with GPT and Make

In 2018, I wrote about using personal health data to generate self-knowledge. Back then, just visualizing my steps and sleep trends was insightful. But in 2025, with the power of GPT-4 and tools like Make, we can go far beyond dashboards.

Today, I use a GPT-powered agent that interprets my wearable data daily and gives me a motivational recovery and activity summary—like a personalized coach that reads my biometrics and gives me tips on how to approach the day.

The Agent

My agent connects data from Fitbit, WHOOP, and Withings (smart scale), transforms it into a readable summary, and then feeds it to GPT-4 via the OpenAI API. For these daily tips, I prompted GPT-4 to keep the tone warm, grounded, and supportive—like a mindful coach or caring friend. Each morning, I get a short, encouraging analysis like:

You’re doing an amazing job balancing recovery and activity—great work getting 9.2 hours of sleep last night and bouncing back with a strong 74 recovery score today after a low-strain rest day on 5/26. That drop in RHR and boost in HRV yesterday shows your body appreciated the break. To support recovery today, keep hydration steady and consider a short breathwork or stretching session mid-day to help regulate your nervous system. For movement, aim for a Light-to-Moderate strain day (6–9)—a brisk walk, easy cycling, or light bodyweight circuit could be perfect. You’re listening well to your body—keep riding that wave of mindful momentum.

I built this without writing traditional code, using Make—a powerful no-code tool I would recommend (and yes, that’s an affiliate link).

Here’s what part of the Make scenario looks like behind the scenes:

Why This Matters

Health device data is often trapped in silos, making it hard to get more holistic coaching tips. While dashboards are useful, they rarely talk to you. This project shows how easy it’s becoming to:

• Unify your data from multiple sources
• Use GPT to make sense of it
• Deliver a coaching experience that adapts to your recovery and activity trends

We’re entering a new phase where agents act not just as interfaces—but as interpreters, guides, and motivators.

What’s Next?

I’m continuing to refine the prompts and logic (e.g., smarter activity and rest day suggestions). Eventually, this could plug into meal/drink logging, strength training cycles, or mood score check-ins.

---

Want the technical walkthrough, error handling tricks, and build instructions? Comment here or message me—I’m happy to share or publish a follow-up if there’s interest.

📺 Watch the Demo

---

👋 If you’re working in health tech or building agentic tools, let’s connect. The agent future is here—and it’s personal.