A workspace that reads the whole chart — not the last ten minutes.
MedScrub connects to your EHR over FHIR and works from the real record — labs, diagnoses, coverage, prior auths. Notes, screening and documentation, all grounded in data, all on your own infrastructure.
Desktop & mobile · HIPAA compliant
The economics quietly work.
Compliance isn't just safe — it's paid for by the workflows it enables.
$149
per month, solo
flat rate
99214
one correctly coded visit
covers the month
11+
HEDIS/MIPS measures tracked
across your panel
0
PHI records leaving your infra
de-identified on-device
Workflows you can't build from a transcript.
Freed, Heidi, Nabla, and Suki start from a recording of your conversation. MedScrub starts from the clinical record — the actual labs, diagnoses, and coverage data.
Population health screening
11+ HEDIS/MIPS/USPSTF measures checked automatically across your panel — before patients walk out.
Prior authorization prep
Reads payer requirements and tells you what documentation you need before you submit.
SOAP notes from the chart
Structured notes grounded in real labs and diagnoses, with ICD-10 and E/M coding support.
CCM documentation
Generate the documentation CPT 99490/99491 requires — from live EHR data, not memory.
Clinical Data Repository
Persistent patient context so the model works from the whole chart, not a single prompt.
Care team, one workspace
HIPAA-safe messaging for MDs, MAs, and front desk — stored on your own proxy.
Four jobs, one grounded workspace.
Start the day with a briefing built from your actual panel — not a transcript of the last ten minutes.
See every open care gap before the visit.
HEDIS, MIPS, and USPSTF measures checked across your whole panel — ranked so the highest-value gaps surface first.
- Colorectal, breast, cervical, and lung screening eligibility resolved automatically
- Evidence pulled from structured FHIR and unstructured documents alike
Describe the task. MedScrub builds the skill.
Turn a plain-English request into a reusable, grounded workflow — with guardrails and a human-review step built in.
- Pipelines chain steps across the chart
- Citations and review gates before anything writes back
Live FHIR from the system you already run.
Connect Epic, Cerner (Oracle), athenahealth, or eClinicalWorks and sync the record into your own Clinical Data Repository — no rip-and-replace.
- Standard FHIR authorization, no IT project
- Persistent context so every skill works from the whole chart
Connect your EHR. Select a patient. AI works de-identified.
01
Connect your EHR
Authorize MedScrub through Epic or athenahealth. Standard FHIR, no IT department needed.
02
Select a patient
Pick a patient. MedScrub pulls their full clinical record: labs, vitals, meds, imaging, problem list.
03
AI works de-identified
PHI is stripped before data reaches the AI model. The LLM sees clinical context, never patient identity.
04
Results with real data
Output is re-identified with real patient data. Review, approve, write back to your EHR.
We're not replacing your scribe. We start from data.
Workflows no scribe can match
Freed, Heidi, and Suki start from audio — one conversation at a time. MedScrub starts from your actual EHR record. This enables MIPS quality tracking, CCM billing automation, and care gap closure that are structurally impossible for conversation-based scribes.
Any LLM, your choice
GPT-4, Claude, Gemini, Mistral, even local models. Because PHI is stripped first, you are not locked into one vendor's model. As AI improves, you improve.
PHI never reaches the AI
The PHI proxy strips all 18 HIPAA identifiers: deterministic, 100% reversible. Your patient data never touches AI servers in identifiable form.
Your Clinical Data Repository
Your patient data syncs into your own CDR, enrichable from HIE networks. It is not locked in another vendor's black box.
Desktop + mobile + API
Native desktop app, mobile, HTTP API, and MCP server. Use MedScrub however fits your workflow.
CCM billing automation
200 eligible patients × $62/mo CPT 99490 = $148,800/year in new revenue. MedScrub auto-generates monthly CCM documentation, turning a complex billing workflow into a reliable revenue stream.
Built to the HIPAA Safe Harbor method.
All 18 identifier types specified in 45 CFR §164.514(b)(2) are removed before any request reaches a model. The design is auditable, documented, and something a hospital CISO can sign off on.
18 Safe Harbor identifiers removed
Names, dates, MRNs, SSNs, addresses and more stripped per 45 CFR §164.514(b)(2) before any AI request.
Self-hosted by design
Runs as Docker in your environment. PHI never leaves your infrastructure. No BAA with a model vendor required.
Reversible tokenization
Re-identify locally after the model responds — including LLM-modified output. Session key stays on your proxy.
From download to first chat
How it works
Five steps from download to chatting with your patient panel. No IT procurement, no PHI ever leaving your infrastructure.
Download and install
macOS, Windows, or Linux. Normal app, not an enterprise rollout. About 10 minutes.
Add your LLM provider credentials
OpenAI, Anthropic, Google Gemini, or any OpenAI-compatible API. Your keys, your costs, your BAA.
Accept MedScrub on your EHR marketplace
Listed on the Epic, athenahealth, and eClinicalWorks app marketplaces. One-click authorization, no procurement, no IT ticket.
Sync with your EHR
First sync pulls your full patient panel into your local data store. Subsequent syncs are incremental and live.
Chat with patient data and your favorite LLMs
Pre-visit summaries, SOAP notes, prior auth letters, MIPS gaps — PHI scrubbed on the way out, restored on the way back.
Most physicians are running their first real query the same day they install.
Common questions.
How is this different from other physician AI tools?
Most physician AI tools are ambient scribes: they listen to conversations and transcribe. MedScrub reads your EHR data directly — structured FHIR data plus unstructured documents (faxed reports, specialist PDFs). This enables population health screening, risk stratification, and compliance tracking that are structurally impossible for conversation-based scribes. You choose your AI model and pay $149/mo instead of $200–$800+.
What screening programs does MedScrub track?
MedScrub tracks USPSTF-recommended screenings and HEDIS quality measures: colorectal cancer (colonoscopy), breast cancer (mammography), cervical cancer (Pap/HPV), lung cancer (low-dose CT), and more. Complex eligibility criteria are handled automatically — e.g., colorectal screening for patients 34+ with first-degree relative history. Evidence is extracted from both structured EHR data and unstructured documents.
Does patient data reach ChatGPT or Claude?
No. The PHI proxy strips all 18 HIPAA identifiers before anything reaches an AI model. The AI sees de-identified clinical data only. When the response comes back, we re-identify it for you. Deterministic de-identification, 100% reversible.
Which EHR systems do you support?
Epic, Cerner (Oracle), athenahealth, and eClinicalWorks connect over standard FHIR APIs — no custom integration or IT involvement needed. HIE enrichment pulls records via CommonWell, Carequality, and eHealth Exchange.
Is this built for solo practitioners?
Yes — no enterprise contract, no IT department, no 3–6 month setup. Connect your EHR and start using it. Solo plan is $149/mo (or $99/mo annual). Practice plan for 2–10 docs is $109/seat.