What's new
Every change we ship, dated, newest first. Short entries published as they go out rather than saved up for a launch — you should be able to tell at a glance whether somebody is still maintaining this.
This page is the web app. The iPhone, iPad, Android and desktop apps are not in any store yet, so nothing about them appears here until it is something you can install.
- Notes
Exporting a note now finds the note. Export and Export all had never matched a single note on any deployment: they came back empty and reported it as though you had none.
- Sessions
Couples, family and group sessions have their own shape — a roster, consent recorded for each person with one refusal governing the whole room, a note with a shared part and a private part per member, and fees split per member.
- Clinical safety
The clinical alerts board now carries crisis numbers: the ones for the region it can work out, and the international directory when it cannot. They appear only while an alert is active, so the strip never becomes something you learn to scroll past.
- Clients
Day counts in client observations no longer shift depending on the server's timezone.
- Sessions
Join opens the Zoom, Meet or Teams link the appointment actually carries, so a session that moved platforms still opens in the right place.
- Booking
sessova.com is Sessova throughout. The booking page you send clients, and the portal they land in, now carry the right name in the browser tab and on the screen.
- Booking
Signing in with Apple or Google as a therapist reaches your own dashboard instead of filing you as somebody's client. A booking charges the price shown for the session type you picked, an abandoned checkout releases the slot again, and both the data export and account deletion run to completion.
- Payments
A client can no longer book for free while Stripe is still verifying your identity. Stripe switches card charges on immediately and holds payouts back for days, and in that window bookings were being confirmed without anyone ever being asked to pay.
- Practice
Published claims now match the code. The SOC 2 badge says certified hosting rather than a certification of ours, the hosting pages name DigitalOcean because that is where the data is, and the Clinic plan lists only what exists today.
- Practice
A therapist who does not practise CBT can switch the CBT machinery off — seven groups of tools, each with its own switch in settings, enforced on the server rather than merely hidden from view. Notes, clients, calendar, messages, billing, consent, risk, clinical alerts and the safety plan can never be switched off, whatever model you work in.
- Clients
Five places where the dashboard told you something untrue are gone: an Assign button that sent no worksheet, a waitlist offer with nothing behind it, caseload percentages generated at random, an invented weekly schedule, and a demonstration telehealth room linked as though it were a real one.
- Booking
Editing your profile no longer takes your booking page offline. Saving a partial change was quietly switching bookings off, so every link you had already sent started returning a missing page.
- Security
Six holes closed after an adversarial audit — among them a way to take over a client account, a way into a live session without joining it, and a referral queue that showed every waiting client's full name and their own words to every therapist on the platform. An unclaimed request now shows a first name only.
- Practice
Appointment reminders, message notification emails and the no-show signal work. None of the three had ever run once.
- Languages
The web apps are complete in nine languages, with the last gaps filled in every one. Validated instruments are now blocked from machine translation: a scored questionnaire is only ever shown in a wording somebody approved, because a cutoff derived from one language does not survive being paraphrased into another.
- Clinical safety
A client answering item 9 of the PHQ-9 in the portal now raises an alert to their therapist. Client-submitted assessments were arriving without their individual answers, so that item was never read — assessments a therapist administered were unaffected.
- Clinical safety
The scanner that reads client writing for crisis language catches far more of how people actually put it: recall on our test set moved from 48% to 78%, with no false alarms on the benign controls. It is deterministic and never involves an AI model, so a disclosure cannot be softened before a human sees it.
- Notes
An AI-drafted treatment plan is no longer published to the client. It stays a draft until you have read it and published it — a plan a client can read is a clinical communication, and it should come from a person.
Nothing is listed here before it is live. If an entry is on this page, it is running in production and you can go and use it.