For Addiction Clinics · Outpatient & Residential
In development · Pre-CE · Targeting Class I DiGA

Give your waitlist
something to hold onto.

You see them in intake, motivated and ready. Then the wait erodes every gain. Kontexa START is being built as the program your patients take home from the consultation — a thirty-second PZN prescription, full GKV reimbursement, and zero burden on your team. Content and validation protocol are ready; CE marking and the DiGA listing are next. We are recruiting partner clinics for the validation cohort now.

The Promise

Patients arrive at therapy
in better shape than they left your office.

No IT integration

Kontexa runs entirely on the patient device. Nothing to install in the clinic.

No staff training

One PZN code added to the existing prescription workflow. That is the entire learning curve.

No ongoing management

No dashboards. No alerts. No risk scores routed back to your team.

No follow-up burden

The patient uses the program independently. The clinic stays in its lane.

What the Patient Does

Two daily check-ins. One weekly module.
Thirteen weeks.

Morning check-in: scaffolded affect labeling. Evening check-in: a short continued inventory. Weekly modules rotate four evidence-based skills. A recovery-stories audio library covers the 2am moments when nothing else is open.

Urge surfing
Observe a craving without acting
Cognitive restructuring
Catch the thought before the behaviour
Problem-solving
A structured alternative to substance use
Social action
Real human contact on demand
Safety Architecture

Three deterministic safety layers.
No adaptive clinical decisions.

01
State machine

All conversation flow is deterministic. The LLM operates inside fixed states. It cannot make clinical decisions or deviate from prescribed weekly content.

02
Output validation

Every model output passes regex and keyword validation. Crisis keywords trigger immediate national crisis line responses regardless of model behaviour.

03
Semantic classifier

A post-generation safety classifier blocks any output flagged as potentially harmful and replaces it with a safe fallback.

Next

Join the validation cohort.
Help shape the category before it exists.

Talk to the team