01
Live
A narrow panel beside your video call. Readable in one glance. Silence is the normal state.
The problem
Listening · runs on your machine
Notice what matters.
Stay with your client.
Session Signals is a quiet live companion for therapists. It listens alongside you and offers one restrained cue, only when the feeling belongs to your client.
What you see during a session. Generated people and voice, invented words.
The safeguard
First it asks: whose feeling is it?
A client can calmly describe someone else’s anger. Text-only tools often tag the client as angry. Session Signals separates the client’s own emotion from emotions they report in other people, and stays quiet when it isn’t sure.
“My manager was furious on Friday, and honestly, I think it’s my fault.”
furious
Belongs to the manager. Recalled, third-party.
No cue
my fault
Client-owned self-blame, said right now.
Cue eligible
One product, three views
During, after, and across sessions.
02
Review
Recording, transcript and every cue on one timeline. See what you used, skipped or saved, and rate whether each cue was right.
03
Patterns
Across sessions, it highlights distinctive links between feelings and subjects, not a graph where everything connects to everything.
You stay in charge
Every cue is a suggestion. One click to keep or dismiss.
Mark a cue helpful, not useful right now, or save it for review. Your feedback is stored locally and never retrains the model automatically.
Client · now · Self
Reflect the feeling
“I keep thinking I’m the one who messed it up.”
Client-owned self-blame, said in the present. Mark it helpful, not now, or save it.
Privacy by architecture
The session never leaves your machine.
- Local model. Classification runs on your own computer. No cloud API sees the conversation.
- No outside connections. The app only listens on your own machine and blocks requests to the internet.
- Your notes stay yours. Feedback and ratings are saved locally, alongside the session.
- Nothing learns behind your back. Feedback is for evaluation. It does not retrain the model on its own.
1cue on screen, at most
0bytes of session audio sent to the cloud
10,409example therapist responses in the local library
3views: Live, Review, Patterns
Clear about what it is
A companion for attention. Not a clinician.
It is
- A restrained, glanceable cue during the session
- A reviewable timeline after the session
- A way to explore patterns as hypotheses
- An evaluation tool for clinical reviewers
It is not
- A diagnosis or a risk score
- A treatment recommendation engine
- A replacement for your clinical judgement
- A clinical probability. Its confidence is a model signal.
Built with and for
Therapists
A second pair of ears that respects your attention and your judgement.
Clinical training
Students review sessions cue by cue and see what an experienced eye might notice.
Clinical reviewers
Structured tagging and ratings to evaluate the model honestly, before anyone relies on it.
Questions
Frequently asked
What is Session Signals?
A live companion for therapists. During a video session it sits in a small panel beside the call and shows at most one short cue at a time, such as “Reflect the feeling”, only when the feeling belongs to your client. After the session you can review every cue on a timeline.
Does it send my sessions to the cloud?
No. Classification runs on your own computer, and session audio is not sent to cloud AI services. Your feedback on each cue is saved locally too.
Is it a diagnostic tool?
No. It is not a diagnosis, a risk score or a treatment recommendation engine, and it does not replace your clinical judgement. Its confidence is a model signal, not a clinical probability.
How does it avoid flagging someone else’s feelings?
It separates the client’s own emotion from emotions they describe in other people. If a client calmly describes a manager’s anger, no cue appears. When it can’t tell whose feeling it is, it stays quiet.
Which video call apps does it work with?
It is designed to run in its own small window beside the video call app you already use, such as Zoom, Google Meet, Microsoft Teams or a telehealth portal.
Does it learn from my feedback?
Your ratings (helpful, not now, save) are stored locally and used to evaluate the model. They do not retrain it automatically.
Can I use it today?
Session Signals is a research prototype. We are showing private demonstrations to therapists, clinical psychology programs and early partners.
See it with a protected demo session.
We’re showing a private demonstration to therapists, clinical psychology programs, and early partners.