AI care agents that work behind glass.
Hosted agents read your health records, interpret labs, track what you're due for, and take action — with PHI redaction, an immutable audit trail, and human confirmation on every write. Graded A to F, live, so the safety is something you check, not something you're told.
A safe place for AI to touch your health records.
You already have the data — scattered across hospitals, labs, and portals. Care Agents brings it together and lets an AI assistant actually work with it: read it, explain it, flag what matters, and do things on your behalf. The difference is that every action runs through a guardrail layer, so nothing happens off the record.
Over chat
Talk to your agent in Telegram. Ask about your labs, what you're due for, or have it draft a call to your pharmacy — it confirms with you before acting.
In your AI assistant
Connect the agent to Claude, ChatGPT, or any MCP client. Same guardrails, in the assistant you already use — read-only by a scoped token you hold.
Purpose-built personas
A PCP advisor, a pharmacy helper, a scheduler — each a focused agent with its own memory, running on the same audited, guardrailed stack.
The safety is the product.
Six properties sit between every agent and your data. They aren't promises in a policy — they're enforced in code, and a built-in harness grades any deployment A to F so you can verify them yourself.
PHI redaction
Identifiers are stripped before anything leaves the boundary. You choose how much your agent sees.
Immutable audit
Every read and write is written to an append-only trail. Nothing the agent does is invisible.
Human in the loop
Any action that changes something — a message, a call, a record — waits for your explicit confirmation.
Step-up authorization
Writes require a cryptographic, time-boxed token. Reading and acting are held to different bars.
Tenant isolation
Your records live in a space only your credential can reach. One patient can never see another.
Honest disclaimers
Every clinical answer is decision support, never diagnosis — and it says so, every time.
Connect once. Watch it work.
Connect your records
Verify your identity once and link your providers through nationwide health networks. Records flow into a private space that's yours alone.
The agent works behind the glass
It interprets your labs, finds gaps in your care, and drafts actions — every step redacted, audited, and disclaimed as it goes.
You stay the final authority
Nothing is sent, called, or written without your yes. And you can pull the live safety grade any time to confirm the rails are on.
Put an agent on your health records — safely.
Care Agents is in early access. Start a conversation with a hosted agent, or bring the guardrailed tools into the assistant you already use.