# MCP Server

The JondaX MCP server lets AI assistants talk to the JondaX health data transformation engine directly. Point an assistant at a lab report or a device photo and it can upload the file, track processing, and pull back clean, structured, coded results — all through natural language, without you writing any integration code.

## What is MCP?

The [Model Context Protocol](https://modelcontextprotocol.io) (MCP) is an open standard that lets AI applications connect to external tools and data sources. Think of it as a bridge: it gives assistants like Claude, Cursor, and Antigravity a safe, standard way to call real services and act on your behalf.

The JondaX MCP server is the JondaX side of that bridge. It exposes the JondaX processing modules as MCP tools, so an assistant can upload files and retrieve results as part of a conversation. The server handles authentication and API calls for you — the assistant just focuses on the task.

## What the JondaX MCP server does

JondaX exposes both processing modules and its retrieval endpoints as MCP tools:

| Tool | Type | What it does |
|  --- | --- | --- |
| `upload_pathology_scan` | Async | Uploads a lab report or pathology document (PDF, image, and more). JondaX de-identifies, extracts, normalises, and translates the data, then returns an `uploadId` for tracking. |
| `upload_medical_device` | Sync | Uploads an image of a device reading (pulse oximeter, blood pressure monitor, glucometer, etc.) and returns the extracted values immediately. |
| `get_upload_status` | Query | Checks the processing and callback delivery status for an `uploadId`. |
| `get_extracted_results` | Query | Retrieves the structured extracted results for a completed upload (JSON, FHIR JSON, FHIR XML, HL7, CSV, Parquet). |


For full parameters and examples, see the [Tools reference](/docs/mcp/tools). To get connected, follow the [Setup guide](/docs/mcp/setup).

## Common use cases

**Conversational document processing**

Hand a file to your assistant and let it run the pipeline end to end:

- "Upload this lab report and tell me when the results are ready."
- "Extract the readings from this glucometer photo."
- "Pull the FHIR JSON for upload `<uploadId>`."


**Batch triage and monitoring**

- "Check the status of these three uploads and let me know which have finished."
- "Which of my recent uploads failed, and why?"


**Format conversion on demand**

- "Get the results for `<uploadId>` as HL7 v2."
- "Give me the CSV version of that pathology extraction."


**Building AI features into your product**

- Add natural-language document intake to a health app.
- Let support agents process a patient file without leaving the chat.
- Wire JondaX extraction into an internal AI workflow or agent.


## How it works

![How the JondaX MCP server works](/assets/mcp-how-it-works.0b0d30b663f04d22cf82123c806c6baf3096e196bf7e03042ae30049425e55de.a1161040.svg)

The assistant acts as the MCP client and JondaX runs the MCP server. A typical pathology run: the assistant calls `upload_pathology_scan` and gets back an `uploadId`, JondaX processes the file asynchronously, and the assistant retrieves the structured output with `get_extracted_results`.

## Supported output formats

- **FHIR**: FHIR JSON, FHIR XML (Observation, DiagnosticReport)
- **HL7**: HL7 v2
- **Standard data**: JSON, CSV, Parquet
- **Terminology & coding**: LOINC, standard clinical units, normalised ranges


## Compliance

The JondaX MCP server runs against the same JondaX engine as the API, built for healthcare data. HIPAA · GDPR · PDPA · ISO 27001