Capture
Use your phone as a handy portable microphone to capture your conversation as you speak, and take optional photos of referral letters, labs, ECGs, or even physical exam findings.
Pilot · ED clinical workflow
Your AI-assisted clinical scribe.
Spoken visits, structured notes.
MDScribe turns your patient conversations in a busy emergency department into a chart-ready clinical note. Speak as you would normally with the patient and as you reason out loud; the app structures everything into a note you can copy straight into your EMR.
Why MDScribe
Built for the workflow of the emergency department.
Designed alongside ED clinicians. Every choice — language model, identifier scrubbing, editing flow — is shaped by what an actual shift looks like.
No hallucinations
Using robust LLM APIs, the app structures only what you actually said into chief complaint, history, exam, observations, discussion, and recommendations.
Multilingual
Support for English, Spanish, and even Hebrew. Speak one language, write the note in another — language is decoupled from the spoken visit.
Learns your style
Edit any section before exporting; MDScribe learns your preferences from the changes you make to the text, and offers to remember them for future notes.
How it works
From conversation to chart-ready note in five steps.
Defense-in-depth at every stage. No human listens to your audio; no LLM sees identifying text. Scroll to walk through each stage.
01
You record
Speak through the visit as you would with any patient. Add photos of referral letters, labs, ECGs, or exam findings if useful. Your phone is the microphone.
02
Audio is transcribed
Speech is converted to text using robust transcription APIs. No human listens. Latency is roughly 10–20 seconds per minute of audio.
03
Identifiers are scrubbed
A defense-in-depth PHI scanner — regex pass plus an LLM scanner — strips names, IDs, phone numbers, and addresses before any generative model sees the text.
04
Note is generated
Only what you actually said is structured into a chart-ready note — chief complaint, history, exam, observations, discussion, recommendations. No invented findings.
05
You review & edit
Adjust phrasing per section, switch any section to another language if needed, then copy straight into your EMR. MDScribe quietly learns from your edits.
Privacy posture
PHI never reaches the model. Audio is gone when you're done.
Built for regulated clinical use. Every safeguard is technical, not promissory.
Pilot disclaimer
MDScribe is in pilot — no patient identifiers should be spoken aloud during recording. The PHI scanner will block recordings that contain names, MRNs, phone numbers, or addresses by default. Visit numbers are entered as form fields and never sent to the AI.
Defense-in-depth PHI scanning
Every transcript goes through a regex pass + Claude scanner before any LLM sees it. Findings hard-fail by default; you can override per encounter and the override is audit-logged.
EU data residency
Database, file storage, and authentication are hosted in eu-central-1 (Frankfurt). Email delivery via Resend in eu-west-1 (Ireland).
Audio is purged after note generation
Once the transcript is generated and the note succeeds, the raw audio bytes are deleted from storage. Transcript metadata (duration, size) is preserved for audit; the audio itself is gone.
No third-party tracking
No analytics scripts, no advertising pixels, no third-party fonts at runtime. Fonts are self-hosted from the deployment.
Try the pilot.
Access is currently invitation-based for clinicians piloting the tool. If you have an invite, sign in with the email it was sent to. Add MDScribe to your phone's home screen for the fastest workflow during a visit.