Documents you already have
Keep uploaded prescriptions, paper records, test reports and payment receipts together, preserving the original files and their sources.
User-added files today
Health Companion · Mobile & desktop app · In development
Use Apple Intelligence to draft document details on a supported Mac and ask about saved records on a supported iPhone. Both features run on the device.
Your health data. Your choices.
My interest in healthcare grew from seeing the differences between healthcare systems in Korea and the U.S. During my Health Informatics course at Georgia Tech, I learned how a lack of medical information sharing between healthcare providers can lead to preventable harm, such as dangerous drug interactions.
These experiences made me want to help people understand and protect their health: to keep their history when they move or change doctors, manage their children’s records when authorized, and choose what to share. A confirmed medication list could help a doctor check for possible interactions when prescribing a new medicine.
The 21st Century Cures Act and its implementing rules have advanced access to electronic health information. Yet my family’s records remain spread across provider portals. Although Apple Health and ChatGPT Health offer provider connections, the providers my family visits did not appear when we searched for them. Paper records and insurance information add to the gaps.
I want Health Companion to help people bring records from different sources together and understand the history behind them, while keeping their original files and sources available.
In Korea, I was familiar with Drug Utilization Review (DUR), which provides medication safety information to doctors and pharmacists, and a patient service for reviewing DUR-checked dispensing history. A dispensing record alone does not confirm current use. In my course team project, we explored clearer lab explanations, questions based on clinical notes, and potential medication interaction alerts for discussion with clinicians.
I want to help people manage their own health information, find and understand it more easily, and get support in caring for their health.
The current app works with files you add yourself. It stores originals locally, lets you review records and group them into Care Journeys, and offers metadata-based related-record discovery with your confirmation.
Keep uploaded prescriptions, paper records, test reports and payment receipts together, preserving the original files and their sources.
User-added files todayIn the future, you could sign in to a hospital or insurer portal and authorize collection of available records, including clinical notes and insurance claims. AI-assisted navigation and institution-supported FHIR connections are being explored.
Future collectionA development build synced a fictional visit record from a Mac to an Android emulator and an iPhone. The iPhone answered a question about that record on-device and showed its source.
Records from different visits, providers and insurers should be connected so you can understand your health in context. The goal is a useful view of your current situation, including medications you confirm you are taking, while preserving each record’s source and keeping your history with you when you move or change doctors, with separate information and access permissions so authorized parents and guardians can manage their children’s records.
Today, you can review imported records and confirm related records in Care Journeys. A comprehensive health summary and confirmed current medication list are future work; a past prescription or claim alone does not establish what you take today.
When new information arrives, the companion should help you review it alongside your history and update your overall picture with your confirmation. For example, when you receive a new prescription, it could help you share your confirmed medications and discuss possible interactions with your doctor or pharmacist.
This review and medication safety guidance are future work and need reliable clinical references and validation. They do not replace professional medical advice.
You may remember that something was found, yet forget the exact size, test value or prescription. The goal is to ask naturally and find the detail without opening every document.
With help from an AI assistant, I want you to understand unfamiliar medical terms in plain language and find the records behind an answer.
iPhone & Mac (Apple silicon) · Apple Intelligence for document interpretation and record questionsWhat size did that earlier imaging report record?
What was my cholesterol result at that visit?
What dose and frequency were prescribed then?
A future answer should show the date, exact recorded value and units, with the source passage. If a detail is missing or “that visit” is ambiguous, it should say so or ask.
On a supported Mac, Apple Intelligence drafts details from PDFs and photos on-device. You can review and edit details, such as a clinic name or date, before saving. On a supported iPhone, Apple Intelligence answers questions about saved records on-device and links to the records it used.
A conversation can help express an intention without navigating every screen or field. You should also be able to open originals, edit suggestions and select values directly. Saving, analysis and sync status, including uncertainty, should remain visible.
Explain unfamiliar medical terms in plain language and make changes such as weight or cholesterol easier to follow over time, with dates, units and original sources available.
Review, edit or reject suggestions. Saving a file does not mean you have agreed to send it to an AI service. AI can help prepare information and questions for a doctor or pharmacist, but it cannot make clinical decisions for you.
Plain-language medical explanations and trend views are future goals. Models such as Google’s MedGemma are being considered for more complex explanations. Using a cloud model would require your permission to share the relevant material.
I want support to build on the history you have confirmed, the signals you allow, and the goals you choose.
“Have you been sitting and working for a while? If it feels right, would you like a short break?”
Recent wearable signals showing little movement could invite a check-in. The companion should ask about your situation and consider your preferences and documented activity restrictions. Low steps or missing readings alone do not prove that you have been sitting.
Wearable ingestion and these check-ins are not implemented. Scheduled checks depend on mobile OS execution limits and data freshness; they are not continuous medical monitoring. You should be able to pause or decline guidance.
I am exploring optional connections to compatible agents, such as Muse or Hermes, so they can help with the context you choose to share. The goal is to let you decide which information each assistant can access.
Agent connections are a research direction; no protocol or integration has been selected. A cloud agent is not assumed to have direct access to a phone’s private files.
On iPhone and Mac with Apple silicon, ask Siri to open the installed Health Companion app by name. On supported iPhones, say “Ask about my records in Health Companion” to start a record question in the app.
Changing a hospital, phone, storage provider or AI assistant should not mean rebuilding your health history.
Retain source files and their origins. Keep extracted values, user confirmations and AI explanations distinguishable.
The goal is portable records with export, encrypted backup and recovery to storage you choose. These capabilities are future work.
Grant specific access to a provider or assistant. Storing a record elsewhere should not automatically grant permission to interpret or share it.
Each device keeps a local copy of the records it has received. The development build supports manual sync between configured devices.
The app keeps records locally. Supported Apple devices can process records on-device, and you can choose when to sync across devices.
AI assistance is part of the app.
A flexible AI connection can help you add, review and understand your records.
The app keeps the records. Future assistance should show sources and let you review suggestions before they are saved.
Now
- iPhone & Mac (Apple silicon): Apple Intelligence
- Android & Windows: External AI (ChatGPT)
Your records, across your devices.
Keep a local copy. Continue on another authorized device.
Now:
- Local records are organized with Git
- GitHub is used as cloud storage to sync between authorized devices
Encryption, conflict handling, backup and recovery need separate implementation and verification.
The current foundation keeps original files distinct from app metadata and confirmed relationships. Future analysis should keep extracted values and explanations linked to their sources. Export, backup, key recovery and conflict handling each need separate implementation and verification.
The application uses a Kotlin Multiplatform core with platform-specific interfaces. Shared record and relationship rules are separated from UI and platform APIs. Storage, synchronization and external-service boundaries are intended to remain replaceable as the project develops.
The current foundation separates original files from app-managed metadata and relationships. Future extraction and chat should preserve the clinical date, item identity, value, unit and source location. Missing or conflicting evidence must stay visible; a historical prescription is not a current medication instruction.
Future export and migration should retain stable record identifiers, source bytes, provenance, relationships and correction status. Backup, synchronization and external AI processing are separate operations. Encryption, key recovery, conflict handling and restore verification need their own implementation and evidence. Local Git history alone is neither encryption nor an independent backup.
A future collector should report collected, failed and potentially missing items, handle retries and changed documents, and return reauthentication to the user. Access policy should apply at each adapter. Operational diagnostics should make failures visible without copying health-record contents or credentials into logs.
Updated November 17, 2026
User-imported files, local source storage, Care Journeys and metadata-based related-record review with explicit confirmation.
The app uses Apple Intelligence and ChatGPT. GitHub syncs records between authorized devices.
Source-grounded conversational recall, plain-language explanations and careful distinction between past prescriptions and confirmed current use.
The app uses Apple Intelligence and ChatGPT.
AI-guided portal collection, FHIR exchange, history-informed wearable check-ins, portable encrypted storage and optional agent connections.
Feature verification varies by platform. This page will be updated as working features and their verification change.