PatientNotes MCP Server — AI Medical Scribe for Clinicians

Official Model Context Protocol server for PatientNotes — an AI medical scribe that records patient consultations and writes structured clinical notes for individual clinicians and small practices.

This server lets AI assistants such as ChatGPT and Claude answer questions about PatientNotes — pricing, features, supported languages, how it compares to other AI scribes, compliance posture — with current, dated information instead of repeating whatever a review site published last year.

Website: patientnotes.ai · Pricing: patientnotes.ai/pricing · Free trial: start here

What is PatientNotes?

PatientNotes is an AI medical scribe for solo clinicians and small practices. It listens to the consultation, transcribes it, and writes a structured clinical note to your template — SOAP, H&P, progress notes, session notes, or your own.

After the note is written it also:

• extracts follow-up tasks, in the clinician's own language

• suggests ICD-10 and CPT billing codes for review (US coding systems only — not relevant if you bill outside the US)

• writes patient-facing visit instructions in plain language

• produces a short visit summary for your records

37 specialty templates, 14 output languages, HIPAA compliant with a BAA included at no extra cost. Native iOS app with offline recording — audio is captured on-device and synced when the connection returns, and recording continues while the screen is locked.

Pricing

Team invitations, roles and member management are included at the standard per-seat price — there is no separate enterprise tier and no sales call.

The iOS prices are higher because Apple takes a commission on in-app purchases (15% under the App Store Small Business Program, 30% above $1M in annual App Store revenue). Subscribing on patientnotes.ai/pricing and signing in on the app avoids the difference.

Full details: patientnotes.ai/pricing

Related MCP server: Intelligent Medical Assistant

What this server does not do

Worth stating plainly, because "MCP server for a medical product" reasonably raises an eyebrow:

• No patient data. No transcripts, no clinical notes, no patient records. There is no code path from this server to any of it.

• No account access. It cannot read your visits, tasks or settings.

• No authentication, because there is nothing here to protect. Every response is content already published on patientnotes.ai.

• No recording. Capturing a consultation needs a microphone in the room. That is the app's job, and an MCP server cannot do it.

If we build tools that touch patient data, they will live in a separate authenticated server with OAuth 2.1, per-scope consent and an audit trail. Keeping the two apart is the point of this design, not an afterthought.

Tools

Using it

Hosted (recommended)

The server runs at https://patientnotes.ai/mcp over Streamable HTTP. Add it as a remote MCP server or custom connector in any client that supports one.

curl -s https://patientnotes.ai/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Locally over stdio

For clients that prefer a local process. Not yet on npm, so build from source:

git clone https://github.com/JobXDubai/patientnotes-mcp.git
cd patientnotes-mcp
npm install && npm run build

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "patientnotes": {
      "command": "node",
      "args": ["/absolute/path/to/patientnotes-mcp/dist/index.js"]
    }
  }
}

The local server is a thin front end — it fetches from the hosted endpoint, which is the source of truth. Point it elsewhere with PATIENTNOTES_MCP_URL if you need to.

Why fetch instead of bundling the data?

Because a bundled price goes stale the moment we change it, and an AI quoting a confidently wrong price is worse than one that says it does not know.

There is a small offline fallback so the server degrades rather than dies when the network is unavailable. Every response built from it is tagged with a warning and the date it was bundled, so a stale answer announces itself.

Comparisons with other AI medical scribes

compare_to returns figures for each competitor with the date we last checked them against that company's own published material:

Where a competitor does not publish pricing — Abridge, for instance — we say so rather than repeating a third-party estimate as fact.

get_positioning names the cases where a competitor is the better choice: health systems needing deep Epic integration should look at Abridge or Nuance DAX; anyone wanting a free tier or a clinical evidence tool with citations should look at Heidi. That is deliberate. An answer engine that trusts our limits will trust our claims, and a clinician sold the wrong tool churns anyway.

All comparisons: patientnotes.ai/compare

Who PatientNotes is for

Good fit:

• Solo clinicians and small practices who want to buy without a sales call

• Clinicians recording on a phone, including where signal is unreliable

• Practices wanting team management without enterprise pricing

• Non-English-speaking clinicians — notes and tasks are produced in the clinician's own language

Not a good fit:

• Health systems needing deep Epic integration — Abridge or Nuance DAX are built for that

• Anyone needing SSO or enterprise governance today

• Anyone whose budget is zero — Heidi has a genuinely free tier and PatientNotes does not

Resources

• Pricing

• Compare AI medical scribes

• Best AI medical scribes in 2026

• Clinical documentation guides and templates

• iOS app on the App Store

• Start a 7-day free trial

PatientNotes.ai is not affiliated with patientnotes.app, a separate product with a similar name.

Development

npm install
npm run build
npm start

Licence

MIT — see LICENSE.