Which clinical documentation vendors have a public API?

Deepgram, AssemblyAI, Speechmatics, AWS (Transcribe Medical and HealthScribe) and Azure AI Speech publish complete endpoint references you can read without contacting anyone. Suki publishes a full partner developer portal with gated credentials. Twofold markets a developer API with self-serve signup but issues the endpoint reference after signup. Ambience releases documentation on request. Abridge and Microsoft Dragon Copilot publish no developer API at all.

Which medical speech-to-text API has the highest published rate limits?

Deepgram documents up to 150 concurrent streams and 50 concurrent pre-recorded requests per project on pay-as-you-go. AssemblyAI documents 100+ new streaming sessions per minute on paid plans with automatic scaling. Speechmatics publishes 50 concurrent real-time sessions on Pro. Azure defaults to 100 concurrent requests on the standard tier, shared across speech-to-text and translation.

How much does a medical speech-to-text API cost per minute?

Published rates at the last verification date: Deepgram Nova-3 batch $0.0043/min and streaming $0.0077/min list, AssemblyAI $0.21/hour pre-recorded plus a $0.15/hour medical-mode add-on, Speechmatics $0.129/hour on Pro, AWS Transcribe Medical $0.075/min, AWS HealthScribe $0.10/min, Azure standard batch $0.18/hour and real-time $1.00/hour.

Do any of these vendors let me build without talking to sales?

Yes. Deepgram ($200 credit), AssemblyAI ($50 credit), Speechmatics ($100 credit), AWS and Azure all provision self-serve with published prices, and Twofold offers self-serve signup with a free trial. Abridge, Ambience and Dragon Copilot all require a conversation before you can see anything.

Is a per-minute transcription price comparable to a scribe API price?

No, and treating them as comparable is the most expensive mistake in this category. A transcription price buys you words; a scribe price buys you a structured note, template adherence and part of the clinical-safety burden. Compare total cost of ownership, including the engineering you keep.

How current is this directory?

Every endpoint, limit and price was verified against the vendor's own public documentation on 2026-09-04. Prices in this category change without notice, so each listing links its primary source — that source, not this page, is the authority.

Reference · Original dataset

Clinical documentation API directory

What each vendor actually exposes to a developer: base URLs, endpoints, authentication, published rate limits, published prices and the integration path — compiled from vendor documentation only.

Written by Compare Healthcare API Editorial DeskReviewed by Technical ReviewLast reviewed 2026-09-01Rubric v1.0

Short answer

Which AI scribe and medical speech-to-text APIs can a developer actually build on?

Six of the ten vendors in this category publish an endpoint reference a developer can read without asking permission: Deepgram, AssemblyAI, Speechmatics, AWS, Azure and Suki's partner platform. Twofold sells a developer API with self-serve signup but releases the reference after account creation. Ambience releases documentation by email request. Abridge and Microsoft Dragon Copilot publish no developer API — they are applications sold to health systems, not components you can embed.

Cite as: Clinical documentation API directory, Compare Healthcare API, last reviewed 2026-09-01.

Key findings

Only 7 of 10 vendors publish an endpoint reference you can read before talking to anyone. Every one of those is a speech-to-text vendor plus Suki's partner platform — the ambient-scribe application vendors publish none.
Published rate limits are almost exclusively an ASR-vendor habit. Deepgram, AssemblyAI, Speechmatics and Azure all state numbers; not one clinician-facing scribe vendor publishes a concurrency figure, which makes capacity planning a contractual exercise rather than an engineering one.
Per-minute prices differ by more than an order of magnitude across the category — roughly $0.0043/min for batch ASR against $0.10/min for a hyperscaler scribe service — because you are buying different amounts of the documentation problem, not different amounts of transcription.
Two vendors offer a discount or a cheaper tier in exchange for model-improvement rights over your audio. For PHI that trade is usually unavailable, so the headline price is not the price you will pay.
The only medical-specific model tiers in the ASR group are Deepgram's Nova-3 Medical and AssemblyAI's medical mode add-on. Azure publishes no medical SKU at all; clinical vocabulary there is a custom-model project you own.

The directory

One row per vendor. Follow a row through to the full endpoint list, published limits, prices and a step-by-step integration guide.

Vendor	API posture	Endpoints	Published limit	Published price	Self-serve
Twofold Partner API for embedded ambient clinical documentation	Public endpoint reference	6	100 requests/second (burst 200); 429 returns Retry-After	$49 per user per month billed annually; $69 month-to-month	Yes
Suki Partner ambient documentation API	Documented, gated access	6	Not publicly documented	Not publicly disclosed by the vendor	No
Ambience Healthcare Gated integration API	Documented, gated access	4	Not publicly documented	Not publicly disclosed	No
Abridge No public developer API	No public API	—	Not applicable — no public API	Not publicly disclosed — contact sales	No
Nuance Dragon Copilot / DAX Licensed enterprise application	No public API	—	Not applicable — no public API	Not publicly disclosed	No
Deepgram Speech-to-text API (medical model line)	Public endpoint reference	2	Up to 50 concurrent requests per project (Nova-3)	$0.0077/min list (promotional $0.0048/min at time of verification)	Yes
AssemblyAI Speech-to-text API (medical mode add-on)	Public endpoint reference	4	5 new sessions per minute	$0.21/hour	Yes
AWS HealthScribe / Transcribe Medical Cloud clinical speech services	Public endpoint reference	4	Per-account, per-region Service Quotas — adjustable by support request; check the Service Quotas console for current defaults	$0.075/min (batch and streaming), per-second billing	Yes
Azure AI Speech General speech-to-text service	Public endpoint reference	3	1 concurrent request — not adjustable	$1.00 per hour	Yes
Speechmatics Speech-to-text platform (cloud or on-premises)	Public endpoint reference	3	2 concurrent real-time sessions	$0.129 per hour (list, at time of verification)	Yes

Developer surface by vendor, verified 2026-09-04

Every listing, in full

Each vendor page carries the base URL, the endpoint table, authentication, rate limits, prices, data-handling terms and an ordered integration guide with its sources.

Twofold API

A documented Partner API (v2) for embedding Twofold's AI scribe inside your own product: OAuth 2.0 provider linking, patient sync, recording launch, webhooks and note/transcript retrieval. Public docs, a full staging environment and self-serve console access.

Production: https://partner-api.trytwofold.com/v2 · Staging: https://partner-api.staging.trytwofold.com/v2 (fully separate environment)

Suki API

The most completely documented developer portal among the clinician-facing scribe vendors: public REST session API, WebSocket audio streaming, webhooks and a staging quickstart.

Environment-specific (staging documented in the quickstart); production base URL issued to partners

Ambience Healthcare API

An integration API for EHRs and healthcare platforms exists and is described publicly — transcription, note generation and coding intelligence — but the reference itself is released on request.

Not publicly documented — released with documentation access

Abridge API

Delivered as a clinician-facing application embedded in the EHR. No developer documentation, endpoint reference or webhook specification is published for third parties.

None published

Nuance Dragon Copilot / DAX API

Sold as a licensed clinical application through Microsoft and its channel. Microsoft publishes licensing and security documentation but no public developer API reference.

None published

Deepgram API

Fully public speech-to-text API with a medical-tuned Nova-3 model selected by parameter. Endpoints, concurrency limits and per-minute prices are all published.

https://api.deepgram.com (EU: api.eu.deepgram.com · AU: api.au.deepgram.com)

AssemblyAI API

Public speech-to-text API with a medical mode enabled by parameter, published per-hour prices, published streaming rate limits and a public sample BAA.

https://api.assemblyai.com (EU: api.eu.assemblyai.com)

AWS HealthScribe / Transcribe Medical API

Two distinct services in one AWS account: Transcribe Medical for clinical transcription and HealthScribe for encounter summarisation with clinical insights. Prices and quotas are published.

https://transcribe.<region>.amazonaws.com

Azure AI Speech API

General-purpose speech-to-text with fast, batch and real-time modes and customisable vocabulary. There is no medical-specific SKU — clinical terminology is a customisation exercise.

https://<region>.api.cognitive.microsoft.com · realtime: wss://<region>.stt.speech.microsoft.com

Speechmatics API

Batch and real-time speech-to-text with published plan concurrency and the category's clearest self-hosting story — the option that keeps audio inside your own perimeter.

https://asyncbatch.speechmatics.com · realtime: wss://eu.rt.speechmatics.com/v2/ (or global.rt)

Evidence & sources

Standards referenced by the integration guides. Vendor-specific evidence is each vendor's own documentation, linked on its listing page.

[1]Business Associate Contracts — sample provisions
U.S. Department of Health & Human Services · Regulation · Tier A — primary documentation
Supports: The contractual clauses a documentation vendor's BAA must contain.
[2]HIPAA Security Rule
U.S. Department of Health & Human Services · Regulation · Tier A — primary documentation
Supports: Administrative, physical and technical safeguards a vendor handling recorded encounter audio must implement.
[3]HL7 FHIR Release 4 specification
HL7 International · Standard · Tier A — primary documentation
Supports: Resource definitions (DocumentReference, Composition, Encounter, Condition, MedicationRequest) that clinical documentation output must map onto.
[4]Word error rate — measurement definition and limitations
NIST speech recognition evaluation literature · Methodology · Tier B — published methodology
Supports: Why a headline WER figure without a stated corpus, audio condition and reference-transcript protocol is not comparable across vendors.

Source tiers are defined on the methodology page. Outbound links are unaffiliated and carry no commercial relationship.

Frequently asked questions

Which clinical documentation vendors have a public API?: Deepgram, AssemblyAI, Speechmatics, AWS (Transcribe Medical and HealthScribe) and Azure AI Speech publish complete endpoint references you can read without contacting anyone. Suki publishes a full partner developer portal with gated credentials. Twofold markets a developer API with self-serve signup but issues the endpoint reference after signup. Ambience releases documentation on request. Abridge and Microsoft Dragon Copilot publish no developer API at all.
Which medical speech-to-text API has the highest published rate limits?: Deepgram documents up to 150 concurrent streams and 50 concurrent pre-recorded requests per project on pay-as-you-go. AssemblyAI documents 100+ new streaming sessions per minute on paid plans with automatic scaling. Speechmatics publishes 50 concurrent real-time sessions on Pro. Azure defaults to 100 concurrent requests on the standard tier, shared across speech-to-text and translation.
How much does a medical speech-to-text API cost per minute?: Published rates at the last verification date: Deepgram Nova-3 batch $0.0043/min and streaming $0.0077/min list, AssemblyAI $0.21/hour pre-recorded plus a $0.15/hour medical-mode add-on, Speechmatics $0.129/hour on Pro, AWS Transcribe Medical $0.075/min, AWS HealthScribe $0.10/min, Azure standard batch $0.18/hour and real-time $1.00/hour. Model your own volume.
Do any of these vendors let me build without talking to sales?: Yes. Deepgram ($200 credit), AssemblyAI ($50 credit), Speechmatics ($100 credit), AWS and Azure all provision self-serve with published prices, and Twofold offers self-serve signup with a free trial. Abridge, Ambience and Dragon Copilot all require a conversation before you can see anything.
Is a per-minute transcription price comparable to a scribe API price?: No, and treating them as comparable is the most expensive mistake in this category. A transcription price buys you words; a scribe price buys you a structured note, template adherence and part of the clinical-safety burden. Compare total cost of ownership, including the engineering you keep. Build vs buy.
How current is this directory?: Every endpoint, limit and price was verified against the vendor's own public documentation on 2026-09-04. Prices in this category change without notice, so each listing links its primary source — that source, not this page, is the authority.