Heidi API and MCP server Let your AI agent finish the notes after clinic.

Heidi's API is for partners such as EHRs and telehealth platforms, and Heidi runs no MCP server for ChatGPT or Claude (checked 16 Nov 2026). Cotlet runs a hosted Heidi MCP server with 7 tools that your agent uses in your own Heidi account. Every change to a note, letter or code waits for your OK.

  • Free 7-day trial, no card
  • Unofficial, not affiliated with Heidi

Cotlet is an independent service. It is not affiliated with, endorsed by or sponsored by Heidi. Heidi is a trademark of its owner. If Heidi's own integrations cover what you need, use them first.

Heidi Cloud Browser

7 tools Read and write

Clinical documentation

What can your AI agent do in Heidi? The paperwork that is left when the last patient goes home.

  • Go through today's sessions and notes

  • Write a session up as a SOAP note

  • Draft a referral letter or certificate

  • Add ICD-10 and SNOMED codes before claims

  • Ask Heidi Evidence a clinical question

  • Keep snippets and terms consistent

Partner API, no MCP server

Does Heidi have an MCP server or an API? The official Heidi routes, checked 16 November 2026.

Heidi documents an API for partners: EHRs, practice systems and telehealth platforms, with access through the Heidi team. Heidi Connectors point the other way, at an MCP server your own EMR or PMS runs. No Heidi MCP server and no official ChatGPT or Claude connector was found.

  • The Heidi developer documentation offers the Heidi API and Heidi Connectors, and says to contact the team for help.
  • Heidi integrates with EMRs and practice systems such as Epic, Oracle Cerner, Cliniko and Halaxy.
  • Heidi Desktop adds Dictate, which types into any app on your computer; it is outside the web account Cotlet works in.
The Heidi API, working by hand and Cotlet compared
What you need Heidi APIBy handCotlet
Ask ChatGPT or Claude to do the work No MCP server found Copy and paste Hosted MCP server
Start today with your own login Partners, by contact Your login Your login, once
Notes, letters and billing codes Endpoints not verified One session at a time 7 tools, writes after your OK
Run from n8n or a script Inside a partner product Manual only REST, same key

Checked 16 Nov 2026. Sources: Heidi developer documentation · Heidi integrations · Heidi on the App Store · Heidi Dictate · Heidi customisation · Heidi documents from one consultation · Heidi Coding · Heidi SOAP notes

Set up once

How does the Heidi MCP server work? The Heidi MCP server by Cotlet: your agent works in your own Heidi account.

  1. 1Sign in to Heidi

    Once, in a browser we host for you.

  2. 2Connect your AI agent

    One address in its settings.

  3. 3Hand it the write-ups

    Notes, letters and codes, each for your review.

Your agent works in your own signed-in Cloud Browser. Every note edit, document and code waits for your OK. Who can see your session

Any MCP client

How to connect Heidi to your AI agent

  1. 1

    Sign in to Heidi in a Cloud Browser. Open scribe.heidihealth.com in your Cotlet Cloud Browser and sign in with your own Heidi account, by email, Google or Apple.

  2. 2

    Add the Cotlet address. Enter https://cotlet.dev/mcp/v1 as the MCP server or connector URL in your agent.

  3. 3

    Let the agent in. Chat apps open a Cotlet sign-in page. Editors, n8n and scripts send an API key made in the console.

Heidi uses the shared address https://cotlet.dev/mcp/v1 with every other app you turn on, so one connector covers them all. Turned Heidi on after connecting? Refresh the connector and the 7 Heidi tools appear.

ChatGPT
Settings > Apps & Connectors > Create
MCP server URL: https://cotlet.dev/mcp/v1/connect
Authentication: OAuth
Claude
Settings > Connectors > Add custom connector
URL: https://cotlet.dev/mcp/v1/connect

Claude Code:
claude mcp add --transport http cotlet \
  https://cotlet.dev/mcp/v1 \
  --header "Authorization: Bearer $COTLET_API_KEY"
Cursor
// ~/.cursor/mcp.json
{
  "mcpServers": {
    "cotlet": {
      "url": "https://cotlet.dev/mcp/v1",
      "headers": { "Authorization": "Bearer ${env:COTLET_API_KEY}" }
    }
  }
}
Codex
# ~/.codex/config.toml
[mcp_servers.cotlet]
url = "https://cotlet.dev/mcp/v1"
bearer_token_env_var = "COTLET_API_KEY"
n8n
MCP Client Tool (inside an AI Agent)
Transport: HTTP Streamable
URL: https://cotlet.dev/mcp/v1
Auth: Bearer, <your key>
Tool: heidi_read
2 read, 5 write

All 7 Heidi MCP tools

One tool per Heidi workflow. The agent works with sessions Heidi has already captured; recording a consultation stays with you.

Clinical notes3

Start here: the session list is how the other tools find a note. Reading runs at once; edits and new SOAP notes wait for your OK.

  • get_clinical_notes List your Heidi sessions for a day, or read the note Heidi Scribe made from one. read

    Heidi Scribe turns a recorded consultation into a structured note, ready to review and refine. This tool reads those notes; it does not record anything, so the consultation itself stays with you in Heidi. It changes nothing and runs without asking.

    MCP callheidi_read → get_clinical_notes

    REST endpointGET https://cotlet.dev/api/v1/heidi/clinical-notes

    Parameters

    Parameters: get_clinical_notes
    NameTypeDescription
    date string string ISO date of the sessions. Defaults to today.
    session string string One session to read in full. Leave empty to list the day.

    Example response

    {
      "date": "2026-11-16",
      "sessions": [
        {
          "session": "Knee pain review, 09:10",
          "template": "SOAP note",
          "codes": 1
        },
        {
          "session": "Diabetes follow-up, 09:40",
          "template": "Progress note",
          "codes": 2
        },
        {
          "session": "Back pain, new patient, 10:20",
          "template": "SOAP note",
          "codes": 0
        }
      ]
    }
  • generate_soap_note Have Heidi write the note of a session in the SOAP format. writeAsks first

    Heidi generates the note from the session itself, as Subjective, Objective, Assessment and Plan. The agent returns the draft for your review; a note that already exists is shown to you before it is replaced. By default you confirm each call; you can switch that off for this tool.

    MCP callheidi_write → generate_soap_note

    REST endpointPOST https://cotlet.dev/api/v1/heidi/soap-note/generate

    Parameters

    Parameters: generate_soap_note
    NameTypeDescription
    session string required string Heidi session, by the name or time get_clinical_notes returns.

    Example input

    {
      "session": "session"
    }

    Example response

    {
      "session": "Back pain, new patient, 10:20",
      "template": "SOAP note",
      "sections": [
        "Subjective",
        "Objective",
        "Assessment",
        "Plan"
      ],
      "saved": true
    }
  • edit_clinical_note Change the text of one section of a Heidi note. writeAsks first

    Heidi notes are AI drafts that clinicians review and refine. The agent shows you the old and the new text of the section before it saves, and changes nothing else in the note. By default you confirm each call; you can switch that off for this tool.

    MCP callheidi_write → edit_clinical_note

    REST endpointPATCH https://cotlet.dev/api/v1/heidi/clinical-note

    Parameters

    Parameters: edit_clinical_note
    NameTypeDescription
    session string required string Heidi session, by the name or time get_clinical_notes returns.
    section string required string Section of the note, by its heading in Heidi, for example Plan.
    text string required string The new text of that section.

    Example input

    {
      "session": "session",
      "section": "section",
      "text": "text"
    }

    Example response

    {
      "session": "Knee pain review, 09:10",
      "section": "Plan",
      "text": "Physiotherapy referral. Review in six weeks.",
      "saved": true
    }

Documents and billing codes2

Letters and certificates from one consultation, and the ICD-10 or SNOMED codes for the note. Documents stay in Heidi as drafts; the agent never sends them.

  • generate_consult_documents Create a referral letter, medical certificate or other document from one consultation. writeAsks first

    Heidi turns one consultation into several documents and can write for a colleague or in plain language for the patient. The document stays in Heidi as a draft for you to read; the agent never sends it. By default you confirm each call; you can switch that off for this tool.

    MCP callheidi_write → generate_consult_documents

    REST endpointPOST https://cotlet.dev/api/v1/heidi/consult-documents/generate

    Parameters

    Parameters: generate_consult_documents
    NameTypeDescription
    session string required string Heidi session, by the name or time get_clinical_notes returns.
    document string required string What to write, for example referral letter, medical certificate or patient explanation.
    instructions string string Anything to add, for example the recipient or a question for the specialist.

    Example input

    {
      "session": "session",
      "document": "document"
    }

    Example response

    {
      "session": "Knee pain review, 09:10",
      "document": "Referral letter",
      "recipient": "Physiotherapist",
      "saved": true,
      "sent": false
    }
  • add_billing_codes Read Heidi's ICD-10 and SNOMED code suggestions for a note, then add, confirm or replace codes. writeAsks first

    Where Heidi Coding is on for your account, it suggests codes linked to the note, ranked by relevance and marked billable. Heidi does not apply codes by itself; the agent shows the codes and adds them only after your OK. By default you confirm each call; you can switch that off for this tool.

    MCP callheidi_write → add_billing_codes

    REST endpointPOST https://cotlet.dev/api/v1/heidi/billing-codes

    Parameters

    Parameters: add_billing_codes
    NameTypeDescription
    session string required string Heidi session, by the name or time get_clinical_notes returns.
    action string required string One of `suggest`, `add`, `replace`. `suggest` only reads.
    codes array array For `add` and `replace`: each code with its system (ICD-10 or SNOMED).

    Example input

    {
      "session": "Knee pain review, 09:10",
      "action": "add",
      "codes": [
        {
          "system": "ICD-10",
          "code": "M25.561"
        }
      ]
    }

    Example response

    {
      "session": "Knee pain review, 09:10",
      "codes": [
        {
          "system": "ICD-10",
          "code": "M25.561",
          "label": "Pain in right knee",
          "billable": true
        }
      ],
      "saved": true
    }

Evidence and customisation2

Heidi Evidence answers with citations. Snippets and preferred terms shape every note you write next.

  • search_clinical_evidence Ask Heidi Evidence a clinical question and get a summary with its citations. read

    Heidi Evidence answers with a summary grounded in the original literature, inline citations linked to their sources and region-aware results. The agent returns the summary and the cited sources as Heidi shows them. It reads only and runs without asking; the clinical judgement stays yours.

    MCP callheidi_read → search_clinical_evidence

    REST endpointGET https://cotlet.dev/api/v1/heidi/clinical-evidences/search

    Parameters

    Parameters: search_clinical_evidence
    NameTypeDescription
    question string required string The clinical question, in plain words.

    Example input

    {
      "question": "question"
    }

    Example response

    {
      "question": "First-line exercise therapy for knee osteoarthritis",
      "summary": "Summary as Heidi Evidence returns it, with its citations.",
      "citations": [
        {
          "title": "Source title as cited by Heidi",
          "url": "https://example.org/source"
        }
      ]
    }
  • customise_note_templates List or add Heidi snippets and preferred terms, or replace a term across your notes. writeAsks first

    Snippets drop saved phrases into a note with a shortcut, the word library keeps your terms and abbreviations consistent, and a term changed once applies everywhere. Shared snippets and terms reach your whole team, so the agent names what changes before it saves. By default you confirm each call; you can switch that off for this tool.

    MCP callheidi_write → customise_note_templates

    REST endpointPOST https://cotlet.dev/api/v1/heidi/note-templates/customise

    Parameters

    Parameters: customise_note_templates
    NameTypeDescription
    action string required string One of `list`, `add_snippet`, `add_term`, `replace_term`. `list` only reads.
    shortcut string string For `add_snippet`: the trigger, for example /bp.
    text string string The snippet text, or the preferred term.
    replace string string For `replace_term`: the word or phrase to replace.

    Example input

    {
      "action": "action"
    }

    Example response

    {
      "action": "add_snippet",
      "shortcut": "/kneeex",
      "text": "Knee exam: no effusion, full range of motion, ligaments stable.",
      "saved": true
    }
Encrypted, revocable

Who can see my Heidi session, and how do I revoke it?

  • What we store

    Your signed-in Heidi browser profile, AES-256-GCM encrypted, in Finland. Never your password.

  • What your agent gets

    The 7 Heidi tools. The session cookie stays in the Cloud Browser.

  • What the session sees

    The sessions, notes and documents your own Heidi account can open, and nothing more.

  • What no tool touches

    Recording consultations, your EMR connection, Heidi team members and your Heidi plan.

  • If Heidi signs you out

    You sign in again yourself, in the same Cloud Browser.

  • How to cut access

    Turn Heidi off for agents, reset the browser, or sign out of Heidi everywhere.

Cotlet never sees or stores your password: you sign in to Heidi yourself, by email, Google or Apple, and the agent only uses the session that sign-in leaves behind. Clinical notes hold patient health information. Our site scores A+ on Qualys SSL Labs, securityheaders.com and Mozilla Observatory (live results).

5 tools ask first

Approvals, undo and mistakes Nothing in a note changes until you have read it.

  1. 01 Find the session get_clinical_notes lists the day.
  2. 02 Ask Heidi for codes add_billing_codes with suggest only reads.
  3. 03 Show you the codes Code, label and billable flag.
  4. 04 Add after your OK add_billing_codes saves the ones you approve.
Fixing a Heidi write that went wrong
ActionHow to put it right
Note section editCall edit_clinical_note again with the earlier text
Billing codeReplace it with add_billing_codes, or remove it in the note
Snippet or preferred termChange it back with customise_note_templates
Referral letter or certificateA draft in Heidi until you send it: edit or delete it there

All 5 write tools ask first: by default you confirm each call, and you can switch that off per tool. The 2 read tools, notes and Heidi Evidence, run at once. AI-generated notes are drafts for the clinician to review, in Heidi and here.

MCP call
{
  "name": "add_billing_codes",
  "arguments": {
    "session": "Knee pain review, 09:10",
    "action": "add",
    "codes": [
      {
        "system": "ICD-10",
        "code": "M25.561"
      }
    ]
  }
}
REST
curl -X POST https://cotlet.dev/api/v1/heidi/billing-codes \
  -H "Authorization: Bearer $COTLET_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"session":"Knee pain review, 09:10","action":"add","codes":[{"system":"ICD-10","code":"M25.561"}]}'
Response
{
  "session": "Knee pain review, 09:10",
  "codes": [
    {
      "system": "ICD-10",
      "code": "M25.561",
      "label": "Pain in right knee",
      "billable": true
    }
  ],
  "saved": true
}

The example adds one ICD-10 code, M25.561, to the knee pain review after you approved it. A script or n8n flow can make the same call over REST at /api/v1/heidi with the same key and price, in JSON. Failed calls are free.

Pay per call

How much does the Heidi MCP server cost?

About $29 a monthfor one clinician over 20 clinic days, 5,000 credits. Your Heidi plan stays as it is.

Illustrative month, our assumption: successful calls for this workload
JobCallsCredits
Read the day's sessions and notes2 × 20400
SOAP notes for new patients5 × 201,000
Edits to Plan sections2 × 20400
Referral letters and certificates2 × 20400
Billing codes on the day's notes5 × 201,000
Heidi Evidence questions1 × 20200
Snippets and preferred terms10 × 1100
Cloud Browser, 30 days1,500
Total350 calls5,000

The smallest pack, 5,000 credits for $29, covers this month. A clinic with several clinicians uses more calls.

  • Free 7-day trial, no card
  • 10 credits per successful call, MCP or REST
  • 50 credits a day for the Cloud Browser
  • Your Heidi plan and features apply as they are

Credit packs are one-time purchases, not a subscription. The month above is an example built from our own assumptions, not measured customer usage.

Cotlet credit packs: price, credits and calls at 10 credits each
Credits Price, one time Calls at 10 credits Per 1,000 credits
5,000 $29 500 $5.80
15,000 $79 1,500 $5.27
32,000 $149 3,200 $4.66
60,000 $249 6,000 $4.15

Only successful calls use credits; failed calls and timeouts cost nothing. At a balance of 0 calls pause, never a surprise charge. Authenticate with the Authorization: Bearer header.

Other options

Heidi MCP server: other ways to connect What each route does for a clinic, and where it stops.

Other ways to work with Heidi, checked 16 Nov 2026
OptionWhat it doesWhere it stops
Heidi APIScribe features inside a partner EHR or telehealth product.For partners, through the Heidi team; you build and host the integration.
Heidi ConnectorsYour EMR or PMS exposes an MCP server and Heidi works inside it.Heidi calls your system; ChatGPT or Claude cannot call Heidi this way.
EMR integrationsHeidi connects to systems such as Epic, Oracle Cerner, Cliniko and Halaxy.Links Heidi to your patient record, not a route for an AI agent.
By handOpen each session in Heidi and work through it.One session at a time, only while you are at the desk.

If you build a clinical product and Heidi accepts you as a partner, its API is the vendor's own route. Cotlet fits a clinician or clinic that wants ChatGPT, Claude or a script working in its own Heidi account today, with nothing to host.

EU company, since 2019

Who runs Cotlet A company you can look up.

Cotlet is built by Take Group sp. z o.o., a software company registered in Poland, EU (VAT PL6342984273), working since 2019 and led by Bart Krawczyk (Co-founder, CTO) and Jake Slugocki (Co-founder, SEO, sales and marketing). Questions about the Heidi tools go to contact@cotlet.dev.

Heidi MCP server: questions

How do I get access?

Sign up free at https://cotlet.dev/signup?app=heidi, turn on Heidi under Integrations in the console and sign in at https://scribe.heidihealth.com/ in your Cloud Browser. Then connect your AI agent and ask it about today's sessions. The trial needs no card.

Does Heidi have an API?

Heidi's API is for partners: its developer documentation describes integrations for EHRs, practice systems and telehealth platforms, with access through the Heidi team (checked 16 Nov 2026). Heidi Connectors let your own EMR or PMS run a small MCP server for Heidi; no Heidi MCP server or official ChatGPT or Claude connector was found. Cotlet does not use that API: it works in your own signed-in Heidi account.

How is this different from Heidi Connectors?

Heidi Connectors work the other way: your EMR or practice system exposes a small MCP server and Heidi works inside it. Cotlet lets ChatGPT, Claude or a script work inside your Heidi account.

Can the agent record a consultation?

No. Recording stays with you in Heidi, on your own device. The agent works with sessions and notes that are already in your Heidi account.

Is the MCP server different from the REST API?

Only in how it is called. Both carry the same 7 tools, inputs, answers, API key and price. Agents use MCP; scripts and n8n HTTP steps use REST.

Is Cotlet affiliated with Heidi?

No. Cotlet is independent and not affiliated with Heidi or Heidi Health.

Hand tonight's Heidi notes to your agent.

Free 7-day trial with 2,000 credits, no card needed.