Your notes, in your approach, signed off by you.

Cogent transcribes the session in your browser and masks identifying details before anything reaches the AI. You get back a note in the approach you used, whether that is CBT, EMDR, ACT or schema, ready to review and sign off.

Watch a session become a finished note.

app.cogent.clinic / clients / sara-w / session-12
Sara W./Session 12/CBT Live 0:00
Live transcription on-device masking · UK-hosted
Capturing
Masked, on your device
Sara [PERSON_1] 943 476 5919 [NHS_1] Lucy [PERSON_2]

The mapping back to the real names is stored encrypted at rest in the UK under a managed key, with access tightly controlled and logged.

Drafting

UK-hosted AI, drafting in your modality's conventions, with nothing in your content ever used to train a model.

Structuring your note
Drafted for your review

Session 12, CBT

12 May 2026 · 50 min · in person
Presenting

Sara returned to the formulation from session 6, with renewed engagement around social withdrawal.

Therapeutic work

Reviewed last week's behavioural experiment. Socratic dialogue around the prediction "if I cancel, they will think less of me".

Risk & plan

No active risk indicators. Continue current CBT formulation, six further sessions agreed.

01Eight modalities

A CBT note should not look like an EMDR note pretending.

You should not have to teach a documentation tool that "client" is the default term, that risk belongs in every entry, that defusion is not restructuring, that limited reparenting is not generic warmth, or that an EMDR note carries phase and SUDS. Cogent holds those conventions the way a supervisor would expect to find them, so the note comes back already in the shape of the modality you used, and the difference is easiest to see with CBT and EMDR side by side.

CBT

Cognitive Behavioural

  • Agenda
  • Cognitive work distinct from behavioural
  • Behavioural experiments reviewed and set
  • Homework
  • Risk in every entry
EMDR

Eye Movement Desensitisation and Reprocessing

  • Phase (1–8)
  • Target memory
  • SUDS and VOC, before and after
  • BLS sets
  • Channels of association

CFT, ACT, IFS, schema, integrative, and assessment hold their own conventions the same way. All eight modalities →

02Living formulation

The plan develops with the work.

After each session note you accept, Cogent reads the new content alongside the existing formulation and proposes per-section updates wherever the note suggests something has shifted, presenting each proposal as a diff with a short rationale that names the session, the evidence, and the change being suggested, so that the plan moves only when you accept the move. There are no silent edits, and the formulation reflects what you have already noticed in the work rather than what an algorithm has decided in your absence.

Version history · Sara W.
v.7
21 Mar 2026
Added safety behaviour to maintaining factors.
v.8
04 Apr 2026
Refined goals after motivation shift.
v.9
21 Apr 2026
Behavioural experiment introduced.
v.10
05 May 2026
Predictions tightened.
v.11 · proposed
12 May 2026
Graded exposure, after disconfirmation.
Diff · v.10 → v.11
Goals+ 1 line

Reduce avoidance. Build evidence against catastrophic predictions. Begin graded interpersonal exposure in close relationships.

Interventions~ revised

Maintain CBT focus. Consider compassion-focused work. Sequence: continue CBT consolidation; revisit CFT at plan-and-review (session 18).

Next stepsunchanged

Six further sessions. Plan-and-review at session 18.

03De-identification in the browser

Real names never reach the AI.

Eight categories of identifying content are detected and replaced with placeholders in your browser, and you confirm the swap, before anything is sent for drafting, so the AI model and the sub-processors only ever see de-identified text. The placeholder mapping that turns the placeholders back into real names is stored encrypted at rest under a managed key on UK infrastructure, with access controlled and logged.

Raw on device Masked, sent to model Restored on return 0 detected 0 sent in clear 0 names to the AI
On your device
Sara W. attended session 12. Reviewed last week's experiment with her partner Lucy. NHS number 943 476 5919. Lives in G77 5RR. Email s.wilson@hotmail.co.uk. Next session 19 May 2026.
Sent to UK-hosted model
PERSON_1 attended session 12. Reviewed last week's experiment with her partner PERSON_2. NHS number NHS_1. Lives in POSTCODE_1. Email EMAIL_1. Next session DATE_1.
04The trust foundation

Three commitments about your clients' material, written into how the product works rather than into a policy document.

01.

You are the author of record.

Every output Cogent produces is a draft you review and accept, with the AI itself never diagnosing, never recommending treatment, never predicting risk beyond what you have already documented, and never interpreting test scores on its own initiative.

Refuses diagnosisalways Refuses risk predictionalways Sign-off requiredevery draft
02.

Identifiers never leave your device.

Beyond the masking demonstrated above, Cogent does not hold the key that would let it read the names, could not be compelled to produce them, and the version of the data stored on Cogent's own servers carries placeholders rather than real identifiers.

Categories detected8 Mapping locationdevice only Encryptionpassword-derived
03.

UK-hosted, with no training on your content.

All hosting, storage, and AI processing happens inside the United Kingdom, the AI provider is contractually prohibited from using your content to train any model, and every sub-processor that touches any part of the work is named in the Trust Centre so the chain can be verified rather than taken on trust.

RegionLondon (UK) Training opt-outcontractual Sub-processorsfully listed
05Comparison

Six dimensions where Cogent diverges from the AI scribes you have already evaluated.

General medical scribes
Generic AI note tools
Cogent Clinic
Built for
The medical encounter, adapted for therapy
General writing
Psychological practice only
Modality conventions (CBT, EMDR, schema, ACT…)
Template library
Generic summaries
8 modalities built in
Identifying details reaching the AI
Sent to the AI on their servers
Sent to global AI APIs
Removed in your browser first
Longitudinal formulation
Living, versioned
Refuses diagnosis and risk prediction
A hard rule
Training on your content
Policy varies by vendor
Often opt-out
Contractually barred
06What Cogent does

Four features carry most of the day-to-day work, and each is built for the way psychology runs.

Those four are the client folder that holds one person's whole record, the transcription that runs in your browser, the pre-supervision brief drawn from the work itself, and the reflective chat that stays with one client at a time, and several more (psychometric scoring, custom templates, a general clinical-knowledge chat, and preparation for the next session) sit a click away on the features page.

  1. Longitudinal client folders
    Quick switch

    Each client gets a folder that fills with the record as the work goes on.

    Drafts, session transcripts, a four-section living formulation, and a thinking-out-loud chat scoped to that one client all sit inside that one folder, with the AI never reading across folders, so the record builds itself as you work and stays in the place where you would already be looking for it.

    How client folders work
  2. Live in-browser transcription
    Start capture

    Sessions are transcribed in the UK with the client's consent and stored encrypted at rest under a managed key, with access tightly controlled and logged.

    You can dictate the note directly or transcribe a session live, with the audio going from your browser direct to an EU-based transcription service in Dublin under the UK Addendum to the EU SCCs, never travelling through Cogent's own servers, and the transcript coming back as text, stored encrypted at rest in the UK under a managed key with access tightly controlled and logged, until you delete it.

    How transcription stays private
  3. Pre-supervision briefs
    Generate brief

    You walk into supervision already knowing what you want to raise, rather than scrambling through the file in the waiting room.

    A short brief drawn from the client's living formulation and the most recent sessions, with case snapshot, current trajectory, present focus, and the questions you want to bring all sitting in one document that opens in the order you would say it in the room.

    • Snapshot of where the work is, rather than a recap of recent events
    • The questions worth raising, pulled out rather than buried in the diary
    • De-identified by default before anything is shared with your supervisor
    Read more about supervision briefs
  4. Reflective thinking chat
    Open chat

    Every client folder has a thinking-out-loud chat that stays with that one client and replies in the voice of peer consultation.

    Every client folder on the Professional tier has a chat that draws on that one client's plan and recent drafts only, so you can compare frameworks, think through sequencing, or sit with a dilemma without receiving directive advice and without anything ever being cross-referenced against another client's material.

    • Anchored to that one client's plan and recent drafts
    • Replies in the voice of peer consultation rather than directive advice
    • Refuses diagnosis, treatment recommendations, and risk prediction
    Read more about the reflective chat

Try it on this week's work.

The free tier covers ten generations a month with no card to start and no time limit on the trial, so you only upgrade when Cogent has earned at least the cost of a session in time saved.

Start free
No card · no demo call · no waitlist