Abridge MCP. Convert audio recordings to structured medical notes.
Gemini Works with every AI agent you already use
…and any MCP-compatible client
Just plug in your AI agents and start using Vinkius.
Abridge (AI Clinical Documentation API) MCP Server automates the entire process of converting spoken patient-provider encounters into structured medical notes.
Your agent handles recording sessions, tracks processing status, and retrieves standardized clinical documentation, including HPI, Physical Exam, and Assessment & Plan sections.
It's designed to integrate directly into clinical workflows, reducing manual documentation time and improving record accuracy.
What your AI agents can do
Create recording
Starts a new recording session or uploads an existing audio file for processing.
Get clinical notes
Pulls structured data containing the summary, HPI, physical exam, and assessment plan for a recording ID.
Get recording status
Checks the current status of a recording task, returning values like pending, processing, completed, or failed.
Starts new audio recording sessions or uploads existing audio files for processing using context data.
Retrieves the current status of a recording task, confirming if notes are pending, processing, or ready.
Retrieves the final, structured clinical documentation, including HPI, Physical Exam, and Assessment & Plan sections.
Ask AI about this MCP
Supported MCP Clients
Gemini Waiting for input…
Abridge (AI Clinical Documentation API) MCP Server: 3 Tools
Use these tools to manage the full workflow: create recordings, track status, and pull structured clinical notes from audio files.
019e5cf6create recording
Starts a new recording session or uploads an existing audio file for processing.
019e5cf6get clinical notes
Pulls structured data containing the summary, HPI, physical exam, and assessment plan for a recording ID.
019e5cf6get recording status
Checks the current status of a recording task, returning values like pending, processing, completed, or failed.
Choose How to Get Started
Build a custom MCP for your own tools, or connect a ready-made integration from our catalog.
Build Your Own
Turn any API into an MCP. Import a spec, define Agent Skills, or deploy with MCPFusion.
- Import from OpenAPI, Swagger, or YAML specs
- Create Agent Skills with progressive disclosure
- Deploy to edge with MCPFusion framework
- Built in DLP, auth, and compliance on every call
- Real time usage dashboard and cost metering
- Publish to catalog or keep private
Make Your AI Do More
Start with Abridge (AI Clinical Documentation API), then connect any of our 4,700+ other servers whenever your AI needs more. One click, no limits.
- Use this MCP plus 4,700+ others, all in one place
- Add new capabilities to your AI anytime you want
- Every connection is secured and compliant automatically
- Track usage and costs across all your servers
- Works with Claude, ChatGPT, Cursor, and more
- New servers added to the catalog every week
What you can do with this MCP connector
Your AI agent handles the whole shebang—turning raw patient-provider conversations into structured medical notes. It's built to plug right into your clinical workflow. You'll use this server to manage the entire cycle, from recording the session to pulling the final, organized documentation.
Manage Recordings
create_recording: Starts a new recording session or uploads an existing audio file. You just need to give it the patient and provider context data.
Check Processing Status
get_recording_status: Checks where the recording task stands. You'll get back a status—like pending, processing, completed, or failed—so you know exactly when the notes are ready.
Extract Structured Notes
get_clinical_notes: Pulls the final, structured clinical documentation. This output includes specific, actionable sections: History of Present Illness (HPI), Physical Exam, and Assessment & Plan.
How Abridge MCP Works
- 1 Your AI client calls
create_recording, providing the audio file and necessary patient/provider context to start the job. - 2 The agent then calls
get_recording_statusrepeatedly, using the resulting ID, until the status is 'completed'. - 3 Finally, the agent calls
get_clinical_noteswith the ID to pull the structured data package containing the full clinical report.
The bottom line is: your agent takes raw audio, starts the process, waits for completion, and then pulls the final, structured medical document.
Who Is Abridge MCP For?
Clinicians who spend time on administrative tasks need this. It's for the medical scribe tired of typing notes after a long day. It's for the health tech developer building an EHR that needs accurate, structured data. It lets you focus on care, not paperwork.
Uses the system to automate note-taking during or after patient encounters, cutting down on administrative time.
Uses the AI-generated drafts as a starting point for high-accuracy clinical records, significantly speeding up the documentation process.
Integrates clinical documentation capabilities into custom EHR or telehealth platforms using the structured output.
What Changes When You Connect
- Saves time on note-taking. Instead of manual dictation, your agent runs
create_recordingon an audio file and handles the rest of the documentation. - Guarantees structured data. The
get_clinical_notestool returns specific sections like HPI and Assessment & Plan, making the output reliable for billing and records. - Manages complex state. You don't have to guess if the notes are ready.
get_recording_statustells you exactly when the AI process finishes. - Reduces manual effort. By automating the note generation lifecycle, you cut down on the administrative burden faced by providers after every patient encounter.
- High data fidelity. The structured nature of the notes ensures key clinical data points aren't missed, which is critical for regulatory compliance.
- Single workflow integration. You can chain
create_recording->get_recording_status->get_clinical_notesinto a single, reliable agent workflow.
Real-World Use Cases
Post-Encounter Documentation
A provider finishes a patient visit. Instead of spending 30 minutes typing notes, they tell their agent to run create_recording on the audio file. The agent then uses get_recording_status and finally get_clinical_notes to pull the complete, structured record instantly.
Integrating into an EHR
A health tech developer needs to build a new module. They use the API to first call create_recording and then use get_clinical_notes to pull structured data directly into the system's database, bypassing manual data entry points.
Reviewing Archived Encounters
A medical scribe needs to review old patient files. They feed the audio into create_recording, monitor the status with get_recording_status, and retrieve the clean, structured notes via get_clinical_notes for quick review.
Batch Processing Multiple Visits
An admin needs to process 10 recordings from a day's worth of visits. The agent loops through the files, calling create_recording for each, and then systematically checks status and retrieves notes using get_recording_status and get_clinical_notes.
The Tradeoffs
Calling notes repeatedly
The agent calls get_clinical_notes right after create_recording and assumes the data is ready. It gets an empty or incomplete response and fails.
→
Always check the status first. Run create_recording to start the job, then use get_recording_status until the status is 'completed'. Only then should you call get_clinical_notes.
Ignoring context
Attempting to create a recording without providing the patient and provider context. The system accepts the audio but the resulting notes are vague and unusable.
→
When calling create_recording, always include the patient and provider context. This ensures the resulting documentation is correctly framed and medically useful.
Stopping at status check
The agent calls get_recording_status and sees 'processing.' It assumes the job is done and tries to retrieve notes, failing with a 'not found' error.
→
Treat get_recording_status as an intermediate step. It tells you when to call get_clinical_notes, but it doesn't provide the notes themselves.
When It Fits, When It Doesn't
Use this if you need to automate the entire, multi-step process of turning spoken word into structured, compliant medical text. You need to manage state: record, wait, retrieve. The workflow is sequential and critical.
Don't use this if you just need to transcribe audio into plain text. For simple transcription, a dedicated transcription API is faster. Also, don't use it if you only need to check the status of a single, pre-existing job; that job should be managed by a simpler status-checking tool.
This server is for high-stakes, multi-stage data transformation where the final output must be structured enough for EHR ingestion.
Independent Platform Disclaimer: Vinkius is an independent platform and is not affiliated with, endorsed by, sponsored by, verified by, or otherwise authorized by Abridge. All third-party trademarks, logos, and brand names are the property of their respective owners. Their use on this website is strictly for informational purposes to identify service compatibility and interoperability.
VINKIUS INFRASTRUCTURE
Cloud Hosted
Managed infra
V8 Isolated
Sandboxed per request
Zero-Trust Proxy
No stored credentials
DLP Enforced
Policy on every call
GDPR Compliant
EU data residency
Token Compression
~60% cost reduction
Works with Claude, ChatGPT, Cursor, and more
The Model Context Protocol standardizes how applications expose capabilities to LLMs. Instead of operating in isolation, your AI gains direct access to external platforms, live data, and real-world actions through secure, standardized connections.
This server provides 3 capabilities that interface natively with Claude, ChatGPT, Cursor, and any MCP client. No middleware. No custom integration required.
Available Capabilities
Documenting patient encounters is a massive chore.
Right now, the workflow is painful. You record the visit, then you have to sit down and manually type notes, cross-referencing what was said with what needs to go into the EHR. You copy a few bullet points, paste them into a template, and then you have to manually structure the HPI, Physical Exam, and Assessment & Plan sections. It's clicking through five different tabs and rewriting everything.
With the Abridge MCP Server, your agent handles the heavy lifting. You just give it the audio file. The server manages the entire conversion process, and you get back a single, structured report with all the required sections—HPI, Physical Exam, etc.—ready for review.
Abridge (AI Clinical Documentation API) MCP Server: Structured Notes
The biggest manual steps that disappear are the initial transcription and the subsequent structuring. You never have to manually categorize a finding as 'HPI' or 'Assessment & Plan' again. The system does it for you.
The result is a clean, ready-to-use document. It’s not just text; it's a structured payload that your system can read and use immediately. That’s the difference between a draft and an actionable record.
Common Questions About Abridge MCP
How do I use the `create_recording` tool with Abridge? +
You call create_recording and provide the audio file and necessary patient/provider context. The tool returns a unique recording ID that you must use for all subsequent calls.
What happens if the notes fail to generate using `get_clinical_notes`? +
If get_clinical_notes fails, it usually means the recording status was not 'completed.' You need to check the status first by calling get_recording_status to find the error or confirm the job is still running.
Can I use Abridge to process live audio streams? +
The current tools are designed for processing recorded or uploaded audio files. You must first use create_recording to establish a session before processing can begin.
Is the structured output from `get_clinical_notes` reliable for billing? +
Yes. The structured output includes standardized sections like HPI and Assessment & Plan, which are designed to align with standard clinical documentation and billing requirements.
How does `get_recording_status` handle long processing times for a recording? +
The tool returns the current status (pending, processing, completed, failed). If processing takes a long time, you can poll the endpoint repeatedly until the status changes to 'completed'. This confirms the notes are ready for retrieval.
What format does the structured data from `get_clinical_notes` use? +
The structured data is delivered in a machine-readable format, designed for easy parsing. It includes specific sections like HPI, Physical Exam, and Assessment & Plan, making it simple to integrate into existing EHR systems.
What happens if the audio file used for `create_recording` is corrupted? +
The system will detect the corrupted file and report an error status. You must correct the file and attempt the recording creation again. The tool does not process malformed audio.
Are there any limitations or rate limits when calling `get_clinical_notes`? +
Rate limits are determined by your Abridge API plan. You should check the Abridge documentation for specific call limits. The API handles concurrent requests, but adherence to plan limits is required.
How do I check if my audio file is still being processed? +
Use the get_recording_status tool with your unique recording ID. It will return whether the task is pending, processing, completed, or failed.
What kind of data is included in the clinical notes? +
The get_clinical_notes tool returns structured medical data including a summary, HPI (History of Present Illness), physical exam findings, and the assessment and plan.
Can I provide patient and provider IDs when creating a recording? +
Yes! The create_recording tool accepts optional patient_id and provider_id strings, as well as a metadata JSON object to provide full context for the encounter.
Use it with your favorite AI tools
Connect this server to Cursor, Claude, VS Code, and more.
More in this category
SMOTE Oversampling Engine
Balance skewed datasets instantly by generating mathematically sound synthetic minority data points via KNN.
Together AI Alternative
Access 100+ open-source models for chat, image generation, and fine-tuning. Power your AI agents with Llama 3.3, Flux, and more.
Chaindesk
Build no-code AI agents trained on your own data that handle customer support, lead qualification, and FAQ resolution.
You might also like
ANVISA (Portal de APIs)
Access official Brazilian health regulatory data — query medicines, prices, food registrations, cosmetics, and sanitizers directly from ANVISA.
Doctolib
Manage medical appointments via Doctolib — search practitioners by specialty and city, track availabilities, and book consultations directly from any AI agent.
PubNub (Real-time Messaging)
Enable real-time messaging, presence tracking, and message history for your AI agent using PubNub's global data stream network.