AI patient intake that knows when to stop.
CareGuard books appointments, captures clean structured intake, and triages symptoms — and hard-stops on red-flag symptoms to send patients straight to emergency care. The safety layer your clinic can actually trust.
No card required · Works with any Postgres · Deploys in a day
Red flags: Chest pain / pressure, Difficulty breathing
Built for GP practices, telehealth, and urgent-care clinics
9
red-flag categories
100%
decisions in auditable code
0
invented appointment times
24/7
intake coverage
What it does
Three jobs, done safely
Two AI agents share one engine — a booking & intake assistant, and a triage assistant with guardrails that can't be prompted away.
Conversational booking
Slot-filling chat collects visit type, date, time, and reason, then books against your real availability. The agent never invents a time it can't honor.
Structured intake
Every conversation is distilled into a clean, typed patient record — chief complaint, symptoms, duration, medications, allergies, conditions — ready for your EHR.
Safety-railed triage
Urgency is decided by fixed clinical rules in code, not the model. Red-flag symptoms force an immediate emergency hard stop. Escalate up when uncertain, never down.
How it works
From first message to safe outcome
Patient chats
A patient opens the assistant and describes what they need — book a visit, or check how urgent a symptom is.
Agent acts with tools
The agent checks real availability, books the slot, and extracts a structured intake record — or runs a rules-based triage assessment.
Guardrails decide
Code — not the model — sets urgency and fires the red-flag hard stop. Every decision is logged for audit.
A hard safety stop you can prove
Most AI health tools bury safety in a prompt and hope. CareGuard puts every urgency decision in auditable code. On a red-flag match, the agent stops the conversation cold and directs the patient to emergency care — no chit-chat, no guessing, no diagnosis.
Explicit red-flag list
Chest pain, trouble breathing, stroke signs, severe bleeding, anaphylaxis, suicidal ideation, and more — each an auditable rule.
Decisions in code, not prompts
The model only extracts symptoms verbatim. A fixed classifier makes the call — so it can't be talked out of escalating.
Never diagnoses
No disease labels, no medications, no doses. Framed everywhere as decision-support and a clinician hand-off.
Audited every time
Each triage decision — and which red flags fired — is recorded, so you can review exactly why the agent acted.
// Red-flag hard stop — overrides everything.
const flags = detectRedFlags(symptomsText);
if (flags.triggered) {
return {
urgency: "EMERGENCY",
redFlag: true,
matchedFlags: flags.matched,
recommendation:
"Call your local emergency number now.",
};
}
// Otherwise: conservative scoring,
// escalate UP when uncertain, never down.Try it now
Talk to the live assistant
Open the assistant in the bottom-right corner. Try both tabs — book a visit, then triage a symptom.
Book & Intake tab
- “I need a telehealth follow-up next week”
- “I've had a sore throat and mild cough for 3 days”
Symptom Triage tab
- “I have a mild headache and a runny nose”
- “I have crushing chest pain and shortness of breath”
The triage tab will hard-stop on that last one — that's the whole point.
Pricing
Simple plans that scale with you
Every plan includes the full safety guardrail layer — it's never an add-on.
Starter
For a single practice getting started.
- 1 clinic location
- Booking + structured intake
- Safety-railed triage
- Up to 500 conversations/mo
- Email support
Practice
For growing multi-provider clinics.
- Up to 5 locations
- Everything in Starter
- EHR export + webhooks
- Up to 5,000 conversations/mo
- Audit log & analytics
- Priority support
Enterprise
For networks with compliance needs.
- Unlimited locations
- Everything in Practice
- SSO & custom red-flag rules
- BAA / HIPAA support
- Dedicated success manager
- SLA & on-prem options
FAQ
Questions, answered
Does CareGuard diagnose patients?+
No. CareGuard provides decision-support only. It never names a confirmed condition, never recommends medications or doses, and always frames itself as a hand-off to a qualified clinician.
How does the red-flag hard stop actually work?+
Triage urgency lives in code, not in a prompt. The model extracts the patient's symptoms verbatim, then a fixed classifier scans an explicit red-flag list. On any match it returns EMERGENCY and the agent immediately directs the patient to emergency care — it can't be talked out of it.
Will it work with our database and EHR?+
CareGuard runs on any standard Postgres — Supabase, Neon, or your own. Structured intake records export cleanly via webhooks and APIs for your EHR on the Practice plan and above.
Can we customize the red-flag rules?+
Yes. The red-flag list and urgency thresholds are configurable, and Enterprise customers can define custom rules reviewed by your clinical team.
Is it HIPAA compliant?+
Enterprise plans include a BAA and HIPAA-aligned controls. This site is a demo; production deployments are configured for your compliance requirements.
Give your patients an assistant that knows its limits.
Deploy safe, guided intake and triage in a day. Keep the clinical judgment where it belongs — with your team.