careagents guardrailed by HealthClaw

Beta: synthetic records only. Things will break, tell us.

Your health, with someone finally on your side of the desk

An agent that actually
knows your health —
and asks before it acts.

Meet your personal care agent. It reads your records through a guardrail layer that redacts what it shouldn't see, writes every access to an audit trail, and hands you the pen before anything real happens.

Get started passkey sign-in · your records stay yours
guardrails: checking… every access audited you approve every action

How it works

  1. 1

    Sign in with your face

    A passkey — Face ID, Touch ID, or your security key — is your login. No password to phish or forget. Then connect your records.

  2. 2

    Ask real questions

    “What do my labs say?” “Am I due for a screening?” Answers come from the records — redacted, interpreted, and grounded, never invented.

  3. 3

    Let it do real work — with your sign-off

    It fills your new-patient intake form from the records. You confirm every medication and allergy yourself — it structurally cannot claim “no known allergies” for you — then you get a provenance-stamped PDF.

Why the guardrails matter

Most health-AI demos ask you to trust them. This one shows its work: the guardrail layer underneath (HealthClaw, open source) grades its own deployment A–F across seven safety properties — redaction, audit, step-up auth, human-in-the-loop, tenant isolation, disclaimers, and truthful failure paths — live, in public. See today’s grade.