Skip to main content
All examples below run on the same Mirobody engine you’ll deploy from Quick Start, with Theta Wellness supplying the polished UI plus some private tools. Chat content uses synthetic health data generated for these demos.
Each showcase below is a real chat session you can open and read in full. Look at:
  • Question — the verbatim prompt the user typed, in plain English (no jargon, no tool names).
  • What the agent does — at a high level, which capabilities of the engine fire to answer it.
  • Open the conversation — the share link to the full transcript with charts.
The eight scenarios target the same three demo personas (a 60-year-old female with cardiovascular history, a middle-aged adult with type-2 diabetes, an adult with anxiety/sleep complaints, etc.) so you can see how one engine reframes the same data for different questions.

​
Private Health Data Management & Chat

Organize scattered data from wearables, healthcare portals, PDFs, and paper reports spanning years into one coherent memory. Then chat over the full context locally or expose it to ChatGPT / Claude via MCP.
The hardest part of personal health data isn’t collecting it — it’s joining “the wearable said X” with “the lab in 2019 said Y” with “the GP in 2024 noted Z”. These two showcases walk through that join from two angles.

​
Knee Pain Progress

Question: “Help me understand the whole picture and the progress of my knee pain, use all relevant data and find potential drivers. Use visual illustrations if it’s helpful.”
Behind the scenes the agent fetches the user’s health profile, searches their indicator history for joint-related markers, pulls events / journal entries / medications, and surfaces a timeline with charts of mechanical-load surrogates (steps, walking pace) layered against pain event dates.

Open the conversation

Knee pain across providers, with synthesized drivers.

​
Cardiovascular History

Question: “What’s the history and recent progress of my cardiovascular issues?”
Same user has 6 years of records spanning two health systems plus device data. The agent cross-references EHR diagnoses, lab trends (lipids, BP), and device data (HRV, resting HR) — and reads the patient profile so the narrative is calibrated to their age, history, and current medications.

Open the conversation

Cardiovascular history across multiple health systems.
Combine device data, medical records, lab work, and self-reported notes into one coherent answer.

​
Cardiovascular Summary

Question: “Summarize all my data related to my cardiovascular conditions across different data sources.”
Same persona as above, different ask: a one-page summary rather than a chronology. The agent pulls vitals, lab series (lipid panel, troponin if present), uploaded reports (e.g. an ultrasound .xlsx parsed via the xlsx skill), and groups everything by clinical theme.

Open the conversation

Cardio summary across sources.

​
Diabetes Tracking

Question: “Summarize the history and recent progress of my diabetes, use data from devices, records, and other relevant sources.”
Type-2 diabetes persona. The agent stitches HbA1c trend, fasting glucose series, medication start/stop events, and the most recent doctor-visit summary — then highlights deltas that matter clinically (e.g. HbA1c trajectory crossing thresholds).

Open the conversation

Diabetes tracking from devices, records, and other sources.

​
PCP Visit Preparation (with linked relative data)

Question: “My father (data linked) just changed his PCP, help prepare a medical history and conditions document that he could bring for his first visit with the new PCP, and potential questions he should discuss with the PCP.”
This one demonstrates multi-user data sharing: the asking user has read access to a relative’s full record via a th_share_relationship row (see API Reference). The agent reads the linked record, produces a print-ready summary doc, and adds a “questions to raise” list tailored to the gaps it notices in the relative’s history.

Open the conversation

Visit-prep doc + question list for a linked relative.

​
Deep Research for Personal Health

Connect your personal data to public knowledge (search, MCP tools, the open web) for deep-research reports — same depth as the agentic research products you’ve seen, but anchored to your data instead of generic guidance.

​
Blood Glucose Analysis

Question: “Help me analyze across different data sources — find out things that help my blood glucose and things that hurt it. Use charts and tables to visualize the findings.”
The agent pivots glucose-series data against food logs, exercise events, sleep duration, and stress markers. Output includes per-factor charts and a ranked “what to keep / what to fix” list.

Open the conversation

Drivers of blood glucose, helpful vs harmful.

​
Symptoms and Feelings Drivers

Question: “Look at my data and analyze the main drivers of my anxiety / stress. Visualize with charts if helpful.”
Anxiety persona. The agent correlates self-reported mood entries with sleep quality, HRV, exercise patterns, and ambient stressors logged in journal entries — then surfaces the top three explanatory factors.

Open the conversation

Main drivers behind symptoms and mood.

​
Diabetes Treatment Options

Question: “Use Theta tools to find out if there are any new methods (new medications, treatments, devices, lifestyle changes, etc.) that could help with my diabetes. I prefer options specifically suited to my situation.”
The agent doesn’t just google “diabetes options” — it grounds the research in the user’s profile (current meds, contraindications, recent lab values, comorbidities) and only surfaces options that match. Sources are cited inline.

Open the conversation

Personalized diabetes treatment options research.
Behind every showcase above is the DeepAgent runtime — single-model, tool-orchestrating, file-aware. The reason that one agent type carries all eight is that personal-health questions almost always require: read profile → search indicator dictionary → fetch the relevant time series → optionally pull files / events / medications → write a synthesis. That’s DeepAgent’s home territory. For cost-sensitive deployments the same workflows run under MixAgent with a cheaper responder phase.

​
Developer Applications

For developers, Mirobody is the data + AI backend you don’t have to write. Focus on the unique parts of your product; let Mirobody handle ingestion, normalization, agent runtime, MCP, and storage.

Wearable Manufacturers

Integrate your device once and instantly offer AI chat over device data.

Research Applications

Drop custom tools into tools/ and SKILL.md playbooks into skills/ to deploy subject-facing research apps without rebuilding the backend.

Consumer Health Apps

Build consumer apps on a complete data + AI engine.

Enterprise

Custom integrations and enterprise support — talk to us.

​
Beyond Health

The same Mirobody architecture works for any “personal data × AI” vertical:

Finance Analyzer

Connect bank accounts, brokerage data, and tax records via custom providers; ask the agent about your portfolio.

Legal Assistant

Drop case files into files/, write a SKILL.md for your jurisdiction, and let the agent draft and cite.

DevOps Copilot

Pipe in metrics, logs, and configs; use sandbox code execution for postmortems and remediation plans.
Swap the files in tools/ and skills/ — same engine, new vertical.