Your health data, verified private.
Most health AI asks you to trust a privacy policy. This app runs inside a hardware-sealed enclave that the operator cannot open. You don't have to believe that — you can verify the running code yourself.
How it works
Upload
Export your Apple Health data as JSON. Upload it through the chat interface.
Held inside a TEE
Your raw data sits encrypted inside an Intel TDX enclave. The operator can't read it. The LLM lives outside the enclave (see limits below).
Chat
Ask questions grounded in your own data. The agent surfaces patterns — it does not give medical advice.
What's proven
Live values from this instance. The external verifier independently confirms the running container matches the registered image.
Opens the EigenCloud dashboard for this app. The platform currently shows 0% on most checks — 3 of 6 are "coming soon" platform features and the remaining 3 require registration steps not yet completed. The raw attestation data (app ID, image digest, KMS key) is shown above.
What this does NOT cover
- The model provider can see your prompts. The EigenAI gateway proxies requests to the inference provider (Anthropic). It does not run inference inside a TEE. Your prompts and the model's responses are visible to the provider in transit.
- End-to-end private inference (where the model itself runs inside an enclave) is a v2 goal, not a v1 guarantee.
- Mainnet-alpha allows developers to upgrade deployed code. Full upgrade-resistance ships in a later EigenCompute phase.
- Single KMS node, operated by EigenLabs. No production SLA — this is a preview environment.
The signing key this enclave uses to attest its identity. Only available when running inside the TEE.
(only available inside the TEE)