Live in production · opening pilot clinics

Your clinic’s front desk, run by AI.

Vitta serves your patients on the official WhatsApp API and does the work inside the system: it books, confirms, reschedules, cancels and wins back patients who lapsed — without a person at the front desk for every message.

Serves patients in Portuguese. Staff can always take over. The AI makes no diagnosis or clinical decision.
V
Example Clinichandled by Vitta · online
Hi! I’d like to book a cleaning next week, preferably in the afternoon.
Hi, Marina! I have Tuesday 2:30pm or Thursday 3pm with Dr. Helena. Which do you prefer?
Thursday!
To confirm: cleaning, Thursday 3pm, Dr. Helena. Shall I book it?
Yes!
Booked ✅ I’ll remind you the day before.
What Vitta executed illustrative
✓offer_alternatives → 2 real calendar slots
✓request_confirmation → awaits “yes”
✓create_appointment → validated · idempotency key
✓google_calendar.sync → event created
✓reminder.schedule → D-1

Illustration with fictional data. Command names simplified.

LivePlatform running on production servers with monitoring and alerting.
Official APIWhatsApp Business Platform (Meta), with official number onboarding.
185versioned database migrations and 200+ automated test files.
CNPJ63.094.483/0001-50 — a real company, verifiable at Brazil’s federal revenue service.

The problem

The front desk became the clinic’s bottleneck.

In Brazil, WhatsApp is where patients want to get things done. Every message needs someone to read it, check the calendar, reply and record it. After hours or at peak times the patient waits — or goes to the clinic next door.

Unanswered messages

Requests arrive at night, on weekends and at peak hours.

A calendar that never closes

Manual cancellations and reschedules leave empty slots and cause conflicts.

Patients who never return

Without active outreach, follow-ups and treatments fall through.

The AI that operates

Not a canned-reply chatbot. An agent that executes.

Patients write however they like. Vitta understands, guides the conversation and performs the matching action in the clinic’s system.

Understands

Interprets the natural-language message and identifies what the patient wants.

Guides

Collects what is missing (service, professional, date), offers alternatives and asks for confirmation before acting.

Validates

The system checks the calendar, conflicts and clinic rules. The model proposes; the system decides.

Executes

Creates, reschedules or cancels, syncs Google Calendar and schedules the reminder. All recorded.

The model proposes. The system validates and executes.

That is why the AI does not “make up” slots or double-book: it never writes to the calendar — it asks the system, which holds the rules. Every command passes a state machine and carries an operation key that blocks duplicate execution.

The AI handles on its ownautonomous

  • Respond at any hour, including after hours
  • Book with real service, professional and time
  • Reschedule and cancel, freeing the slot
  • Confirm appointments and send reminders
  • Answer questions from the clinic’s trained knowledge base
  • Reactivate patients who did not return

Staff steps in on demandhuman control

  • Take over any conversation in one click and hand it back later
  • Turn the AI on or off for the whole clinic or for chosen contacts only
  • Automatic handoff when the AI decides a person is needed
  • Caps on autonomous outbound messages
  • All clinical decisions stay with the professional

Beyond patient service

A complete clinic management platform.

The AI does not sit on top of a spreadsheet: it works inside the system where the clinic actually runs. That is what makes it hard to copy.

Scheduling

Multi-professional and multi-room, walk-ins, blocks and Google Calendar sync.

Records & anamnesis

History, notes, prescriptions and link-based digital anamnesis with versioned templates.

Finance & commissions

Payables and receivables, cash flow, contracts and commissions by professional and service.

Management

Dashboard, reports and KPIs; a dedicated view for each professional.

Trust

Built for health data and for running without surprises.

Serving patients demands precision and care with sensitive data. These are controls that exist in the product — not marketing promises.

Per-clinic isolation

Row-level security (RLS) in PostgreSQL, with separate runtime and migration roles.

LGPD end to end

Flow for data-subject requests (access, correction, deletion) and automatic time-limited purge of artifacts.

Access & attachments

Role-based access, optional two-factor (TOTP) and antivirus scanning of attachments.

Replay-safe execution

Operation key plus distributed lock: the same action never runs twice.

Observability

Metrics, proactive alerts and conversation-flow monitoring.

A kill switch

Turning the clinic’s AI off takes over every conversation at once and clears exceptions.

See security & privacy details

Where we are

Honest about the stage.

We would rather show what exists and what comes next than inflate numbers.

Available today
  • AI patient service on the official WhatsApp API
  • Book, reschedule, cancel and confirm
  • Reactivation and handoff to staff
  • Scheduling, patients, anamnesis, documents and finance
Now: pilots
  • Pilot clinics with before/after measurement (no-show, front-desk time, conversion)
  • Evaluating Claude as the reasoning and tool-use layer
  • Case studies with consent
Experimental / next
  • Clinician-reviewable transcription and summary — tested on synthetic data only
  • Expanded clinical assistant, always with human review

Claude

Why Vitta is a natural fit for Claude.

Agents that take real actions in customers’ systems depend on three things: reliable tool use, faithful rule-following and strong Portuguese.

Tool use

Every conversation becomes a sequence of validated commands: offer slots, request confirmation, create, reschedule, cancel, hand off.

Rule-following

In healthcare, a wrong slot or drifting into clinical advice has a cost. We need a model that respects boundaries.

Brazilian Portuguese

Patients write with slang, transcribed audio and abbreviations. Natural replies are part of the clinic’s experience.

Transparency: today Vitta runs on OpenAI models; we do not use Claude in production yet. The plan with the program is to build an evaluation on real flows (book, reschedule, cancel, hand off), compare quality, error rate and cost per conversation, and move what Claude does better — using the “model proposes, system validates” architecture that makes that swap safe.

Company

Who is behind it.

Vitta is a product of OrbySoft, a Brazilian applied-AI startup. Everything below is verifiable.

Company
OrbySoft — orbysoft.com.br
CNPJ
63.094.483/0001-50 · check at the federal revenue service
Legal form
Individual Micro-Entrepreneur (MEI) — a Brazilian legal entity with its own CNPJ
Founded
October 2025
Headquarters
Santo Antônio de Jesus, Bahia, Brasil
Founder
Josué Moura Santos — profile · GitHub
Contact
josuesantos@orbyvitta.com · WhatsApp +55 (75) 98119-8415
Funding
Self-funded (bootstrapped).
Platforms
Official WhatsApp Business Platform API (Meta); Sign in with Google and Google Calendar via a Google-verified app.

FAQ

What people usually ask.

Does the AI really handle patients on its own, with no front-desk staff?

For routine work — booking, confirming, rescheduling, cancelling, answering questions and reactivating patients — yes: the AI runs the conversation and performs the action in the system. The clinic sets how far it goes and can take over any conversation at any time. Anything outside what the AI should handle is passed to a person.

What stops the AI from booking a slot that does not exist?

The language model never writes anything directly. It proposes a command; the system checks the calendar, professional, service and conflicts, and only then executes. Each command carries an operation key that prevents duplicates if the same message arrives twice.

Does the AI make clinical decisions?

No. It handles operations (scheduling, confirmations, relationships). Diagnosis and treatment are always the health professional’s call.

Does it use the official WhatsApp API?

Yes — the official WhatsApp Business Platform API (Meta), with the clinic’s number onboarded through Meta’s official flow. We do not sell unofficial automation.

Which AI models does Vitta use today?

Today the language layer runs on OpenAI models. Because the model only proposes commands that the system validates, the architecture lets us swap or compare models without rebuilding the clinic’s rules — which is what we plan to do with Claude (see the Claude section).

Do you already have clinics using it?

Vitta is in production and we are in the pilot phase: opening seats for partner clinics that want to run with the AI and measure the result with us. We would rather say that plainly than show numbers we cannot yet prove.

Is patient data protected?

Each clinic is isolated inside the database (row-level security in PostgreSQL), there is a flow for data-subject requests under Brazil’s LGPD, automatic purge of temporary files and antivirus scanning of attachments. Details on the security page.

Want to see the AI run your front desk?

We are selecting pilot clinics. Tell us how patient service works today and we will show, in practice, what the AI would take over.