HoldNote
A local-first clinical scribe for therapists and independent clinicians that captures session audio and produces structured notes, treatment-plan updates, and billing suggestions entirely on-device so PHI never leaves the laptop.
Solo therapists and small private-practice clinicians who refuse to send patient audio to cloud AI
- One-click session recording with local-only transcription and PHI redaction for storage
- Auto-drafted SOAP / DAP notes and treatment-plan updates from session audio using on-device models
- Cross-session progress tracker that highlights movement toward goals without transmitting patient data externally
- Air-gapped export — nothing leaves the device unless the clinician explicitly clicks 'send to EHR' or prints
Hardware like the AMD Ryzen AI Halo dev kit makes on-device LLMs viable for a $4k setup, just as cloud-based clinical scribes like Abridge raise HIPAA alarm bells in therapy communities.
Active therapist adoption of AI scribes (Upheal, CliniScripts, Skriber) plus explicit NPR/CBC coverage of privacy concerns shows real pull for a HIPAA-safe option.Therapists say AI can help them help you, but some see privacy concerns ↗Mental health therapists who use AI to take notes face questions ↗
Cloud AI scribe market is saturated and very well-funded ($1.5B invested in 2025; Abridge, Ambience, Freed, Upheal), and no major player ships a local-first mode; the on-device angle is differentiated but window may close fast as incumbents add offline tiers.Healthcare AI Scribe Competitive Analysis: Gaining Strategic Advantage ↗Abridge vs Ambience vs Freed ↗
Abridge hit ~$100M ARR proving therapists/clinicians will pay for scribing, but solo private-practice budgets are tight ($50–150/mo typical) and an on-device product limits recurring-revenue stickiness and shrinks addressable buyers.Abridge vs Ambience vs Freed ↗Upheal | EHR for Therapists ↗
On-device AI is a multi-year tailwind (Apple Silicon, AMD Ryzen AI Halo, Foundry Local, Gemini Nano) and HIPAA scrutiny is intensifying, structurally favoring local-first clinical tools.The On-Device Agent Era: Local AI Goes Personal in 2026 ↗Harnessing Local LLMs for Private, Efficient Medical Transcription ↗
Local STT is proven (LocalWhisper, OpenWhispr on Whisper/Parakeet) and combining it with a small local LLM for SOAP/billing is demonstrable, but a 60-min therapy session with structured notes, treatment plans and billing is heavy, and requiring a ~$4k Ryzen AI Halo box is a hard go-to-market friction for a solo therapist who usually has a laptop.OpenWhispr | Open Source Voice to Text & Dictation App ↗Harnessing Local LLMs for Private, Efficient Medical Transcription ↗