When you ask an AI health app "why am I so tired this week?", something has to happen with your health data. The question is: where does it go, who sees it, and how long do they keep it?

For most AI health apps, the answer involves a server you don't control, an API key the company owns, and a privacy policy that technically permits a lot of things you probably haven't thought about. The Bring Your Own Key model changes all of this — and it's worth understanding why.

The Standard AI Health App Model (and Its Problems)

The conventional architecture looks like this: you open the app, ask a question, the app sends your health data to the company's server, the server calls an AI API (using the company's key), gets a response, and sends it back to you.

Traditional AI health app data flow
📱
Your iPhone
Health data
→
🏢
Company Server
Stores & processes
→
🤖
AI Provider
OpenAI / etc.
Your health data touches the company's server before reaching AI. It may be logged, stored, or used for model training.

The problems with this model for health data specifically:

Storage risk. If the company's server stores your data — and many do, for features like history and personalization — that data becomes a liability. Breaches happen. Companies get acquired. Privacy policies change. Health data is among the most sensitive information that exists.

Opaque usage. What the company does with your health data beyond answering your question is often described only in dense legal language. "Improving our services" and "aggregated insights" can mean a lot of things.

Markup costs. The company pays for the AI API and charges you a margin, often embedded in a subscription. You're paying more per query than if you went direct — and you don't know how much more.

The BYOK Model: A Fundamentally Different Architecture

Bring Your Own Key works differently. Instead of routing your data through a company server, the app uses your personal API key to communicate directly with an AI provider from your device.

BYOK data flow (Metrya's model)
📱
Your iPhone
Health data
→
🤖
Your AI Provider
Direct connection
No intermediate server. Your device talks directly to Anthropic, OpenAI, or Google using your key.

The app itself doesn't operate backend servers. It has no way to see, log, or store the content of your health queries — because they never pass through it. Your data goes from your iPhone directly to the AI provider you've chosen, using the key registered to your account.

What BYOK Changes (and What It Doesn't)

✓ What improves with BYOK
✓ App company has no access to your health queries
✓ No app-side storage of your health data
✓ You choose which AI provider handles your data
✓ You pay AI providers at their published rates — zero markup
✓ You can revoke access by deleting your API key
⚠ What doesn't change
→ Your AI provider still receives and processes your data
→ Their privacy policy applies to your queries
→ You need to trust the AI provider you choose
→ API key management becomes your responsibility

The BYOK model doesn't make your health data invisible to all parties — the AI provider you choose still processes it. What it does is remove one significant intermediary (the app company) from the data chain, and puts you in control of which AI provider handles your information.

Provider options in Metrya: Anthropic (Claude), OpenAI (GPT-4o / o3), and Google (Gemini). Each has its own privacy and data policies. Anthropic's API by default does not use submitted data for training; OpenAI and Google have similar options. Check each provider's current policy for specifics.

The Cost Transparency Benefit

Privacy isn't the only reason BYOK is worth understanding. Cost structure is the other.

When an app owns the AI key and bundles the cost into a subscription, you're paying an opaque markup on the underlying compute. With BYOK, you pay the AI provider directly at their published API rates — typically fractions of a cent per query for most health questions.

ScenarioEstimated cost
"Why is my HRV low this week?" (Claude Sonnet)~$0.002–$0.005
Full morning health brief (2,000 token prompt)~$0.004–$0.01
30 queries per month~$0.06–$0.30
Typical bundled AI health subscription$9.99–$19.99/month

Your actual API costs depend on which model you choose and query length. But the transparency is valuable in itself — you can see exactly what you're spending and adjust accordingly.

Getting Your First API Key

Each provider takes about 3–5 minutes to set up:

Anthropic (Claude)

Go to console.anthropic.com → Create account → API Keys → Create Key. Anthropic offers free trial credits for new accounts. Claude tends to give nuanced, well-caveated health analysis.

OpenAI (GPT-4o / o3)

Go to platform.openai.com → Sign up → API Keys → Create new secret key. OpenAI requires a small minimum credit purchase. GPT-4o is fast and direct.

Google (Gemini)

Go to aistudio.google.com → Get API key. Google currently offers generous free tiers on Gemini. Good for high-volume use.

Once you have a key, paste it into Metrya's Settings screen. It's stored in iOS Keychain — the most secure storage available on iPhone — and never leaves your device.

Why This Model Is Growing

BYOK is appearing across a range of sensitive-data AI applications: legal research tools, personal finance assistants, mental health companions, and now health apps. The pattern is consistent: where data is genuinely sensitive and users are sophisticated enough to care, the BYOK architecture offers a meaningful privacy improvement over the standard model.

For health data specifically — which can affect insurance, employment, and personal relationships in ways that most data cannot — the choice of architecture isn't just a technical detail. It's a question of who ultimately holds information about your body.

Your health data. Your AI key. Your rules.

Metrya uses BYOK so your health data goes directly from your phone to the AI provider you choose — no server in between, no data stored, no markup.

Download Metrya — Free