Thryve x BlueNexus: Building an AI Health Agent with Wearable Data

Most conversations about AI agents in digital health stay quite theoretical: use cases, architecture diagrams, roadmaps. But BlueNexus, a no-code platform for building AI agents, set out to prove the opposite, that a genuinely health-aware agent, one that reasons over real physiological data rather than a scripted flow, can be assembled and deployed in a single hour. Doing that for a mental health coaching agent meant solving a problem most no-code platforms sidestep entirely: getting continuous, reliable wearable data behind the agent without turning the build into an integration project. To close that gap, BlueNexus built its agent on top of Thryve's wearable data infrastructure.

The Partnership

BlueNexus is the agent-building platform: an interview-driven, no-code interface that turns a rough brief into a fully specified agent, with skills, automations, and a live deployment, without the user writing code. As part of a public build series, BlueNexus founder Chris Were used the platform to construct "Lumen," a mental health wellness coach agent, live and end to end, in under an hour.

Thryve's role was to anchor the data layer underneath it. Rather than BlueNexus building or maintaining a wearable integration, the agent connected to a user's Apple Health data through Thryve's Wearable API, giving it a single, ready-made source of sleep, activity, and heart-rate data to reason over from the moment the build started.

The Challenge

Before the build could focus on what makes a coaching agent useful, BlueNexus and Thryve had to account for a set of constraints specific to wearable-aware, no-code agents:

• No margin for integration work: with the entire build capped at one hour, there was no time to write device-specific parsers or manage a syncing pipeline, wearable data needed to be available the moment the agent asked for it.
• Heterogeneous signal types: sleep, activity, heart rate, and (potentially) HRV each behave differently and needed to be usable by the agent's reasoning without custom handling for each metric.
• Health data sensitivity: an agent reading continuous biometric data and taking action on it, nudges, escalations, care suggestions, needed a data layer built with consent and privacy handling in mind, not bolted on afterward.
• Reliability under agent control: the agent needed to trigger data pulls on its own schedule (daily check-ins, on-demand trend analysis) rather than relying on a human to manually refresh a dashboard.
• Keeping the no-code promise intact: the wearable connection had to be simple enough that a non-developer building an agent through BlueNexus's interview process could wire it up without ever touching an SDK or API key.

The Solution: A Wearable-Aware Agent Built on Thryve's Infrastructure

Thryve provided the digital health infrastructure underneath Lumen, giving the agent a single, always-on connection to wearable data that BlueNexus could build multiple skills on top of, without any custom integration work during the build itself.

• Single connection point: linking Apple Health through Thryve gave the agent one reliable source of wearable data, available to every skill in the agent from the moment it was connected.
• Baseline screening: the agent walked users through PHQ‑9, GAD‑7, and PSS‑10 screening tools, delivered through a purpose-built mini-app, and stored results against the user's profile to establish a starting point.
• Pattern insight skill: a recurring skill pulled the user's wearable data alongside calendar load and screening history, computed a composite wellness score, and looked for signals worth acting on, without a human reviewing the raw feed.
• Escalation skill: when a pattern crossed a concerning threshold, the agent triggered a real, human-in-the-loop email alert, closing the loop from passive data to an actual notification.
• Care navigation: wearable-derived trends were cross-referenced against a benefits and provider database to surface a specific next step, a covered counseling session or a matching provider, rather than just a flag that something looked off.
• Zero custom code: because the data layer was already in place, none of the hour went toward wearable integration, all of it went toward screening logic, escalation rules, and the user experience around them.

The Results

Since the build was completed and published, it has stood as a working example of what BlueNexus's platform and Thryve's data infrastructure can produce together:

• A fully working agent, live: Lumen was built, tested, and published to its own website within the one-hour session, not a mockup or slide deck.
• Three distinct skills, one data connection: baseline screening, trend analysis, and care navigation all ran off the same single wearable link, demonstrating that one integration point can power multiple layers of an agent's reasoning.
• No integration overhead: the entire build hour went into agent logic rather than data plumbing, the kind of trade-off BlueNexus needs to hold at scale for every user building on its platform.
• A public proof point: recorded and streamed live, the build now serves as a concrete demonstration to BlueNexus's audience of what a wearable-aware health agent looks like when the data layer is already solved.
• A template for future builds: the same pattern, one Thryve connection, multiple downstream skills, gives BlueNexus a repeatable foundation for other health-adjacent agents beyond this one use case.

This collaboration shows what's possible when an agent platform doesn't have to solve wearable data infrastructure itself: BlueNexus could focus entirely on making Lumen a genuinely useful coach, while Thryve made sure the data behind it was there, normalized, and ready the moment the agent needed it.