Your business stays under your control.
What we protect and how we monitor it. We only tell you what we can prove.
Your service, always at hand
A website that is down costs you customers. Here is what we do to prevent it.
Continuous monitoring
We check the website, the API and the widget continuously; if anything fails, our team is alerted.
Access with a backup
If the main access fails, we switch to a backup access within minutes.
Protection against attacks
We filter malicious traffic before it reaches your website, your dashboard and the API.
Multiple backups
Every day we keep copies of your data: a main one and a backup.
Voice and messages you can trust
An AI that serves your customers has to know when not to make things up and when to hand over.
An appointment only exists if it really exists
The AI doesn't confirm a booking until the booking system has created it. Automatic checks verify this in every conversation, whether by chat, messages, email or voice.
Prices and hours, from your calendar
Prices and availability come from your business data, not from generated text: the AI checks your real calendar before answering.
A person, whenever you want
Your team can take over any conversation at any time, and calls can be transferred to a person.
AI disclosure and recording, with safeguards
The AI identifies itself as such at the start of the conversations it handles, and that notice cannot be hidden, not even with white label. If you turn on call recording, callers are told at the start, recordings are kept for 90 days and can be exported.
Your information
What we store, how we protect it and who else sees it.
Credentials encrypted at rest
Third-party credentials (WhatsApp, calendar, email) are encrypted at rest and excluded from logs and plain text.
Each business, isolated from the others
Each business's data is isolated with controls in the database, and automated tests verify it on every change before it is released.
The AI never sees third-party credentials
The model interprets the conversation and requests tools through a controlled toolset of our own; it never receives tokens or executes anything directly. Transcript translations are stored separately from the original, which is never modified, and are only generated if the business asks for them or switches on automatic translation.
Who processes your data, and where
We publish which categories of providers process data, what they use it for and in which region, including artificial intelligence.
Protected, idempotent incoming notices
The notices we receive from providers (payments, messaging) are rate-limited and idempotent — a resent event is never processed twice.
Your data is yours
Your customers can ask for access to, correction of or erasure of their data from a public form, and you manage those requests with controlled deadlines from the panel. If you ever leave, you can download all your data and see what is deleted and what is kept.
Data minimisation
Only what is needed to handle enquiries and make bookings is stored. If you connect a clinical integration, it synchronises the patient data that integration needs: ask for the data processing agreement (DPA) before switching it on.
Control and traceability
Who can do what in your account, and how it is recorded.
Roles and two-step verification
Everyone on your team signs in with the role they need — owner, collaborator, reader, billing, chat operator, limited calendar, teacher or read-only auditor. From 3 November 2026, two-step verification is mandatory for owner, billing, collaborator and auditor; the other roles can turn it on whenever they want.
Full audit of every action
Every tool run by the AI is logged — what was called, with what data, with what result — and can be reviewed.
Business activity log
The owner and the Auditor role see a filterable history of what happens in the account — recordings, consents, changes — so that it can be demonstrated in an inspection.
Append-only activity log
The application can only add entries to the activity log: it cannot modify or delete entries already written, and the database itself enforces that restriction. The only exception is anonymising the personal data of someone exercising their right to erasure.
Single sign-on
For large teams, SSO with SAML (Okta, Microsoft Entra, Google Workspace) and a DNS-verified domain. Enterprise only; activated on request.
Every change, reviewed before release
Before a change is released, automated tests and analyses look for exposed secrets and dependencies with known vulnerabilities, and new ones are monitored.
What we don't promise
We don't promise that nothing will ever fail. We promise to monitor, alert and tell you: the real availability history is public on the status page.
See the service status →