NLP API MCP Server

Overview

emtelligent offers a Model Context Protocol (MCP) server that allows AI assistants to use the NLP API medical language engine as a source for medical ontology concept identifiers (codes) for ontologies such as SNOMED-CT or ICD-10.

This is a beta tool. By enabling this app, you acknowledge that you are using an experimental Model Context Protocol server to connect to large language models (LLMs) and the output you see is a combination of the MCP tool output and the LLM output. You may encounter bugs, errors, or unexpected results. This tool is intended for existing emtelligent customers who have API access to the NLP API service.

What is MCP?

Model Context Protocol (MCP) is a standardized way for AI assistants to connect with external data sources and tools. The NLP API MCP Server exposes various tools to the AI assistant. These tools provide codes for concepts in different medical ontologies (like SNOMED-CT or ICD-10). This enables AI assistants to perform reasoning over these ontology codes and to side-step any inaccuracies that might result in the AI assistant hallucinating these ontology codes. The reasoning done by the AI assistant can provide a complex combination of outputs from multiple ontologies to solve many different tasks in healthcare.

Quick Start

Server URL: https://api-beta.us.emtelligent.com/mcp?api-key=<your-api-key>

The emtellipro remote MCP server can be used with AI agent apps such as claude.ai or Claude Desktop.

Requires a Claude Pro account for setting up the MCP server on claude.ai or Claude Desktop.

Go to the “Search and tools” button just under the text box for your prompt and select ”+ Add Connectors” and then “Manage Connectors”. This will take you to the Settings page for Connectors. Select “Add custom connector” on this Settings page.

Select Name to be emtellipro and Remote MCP server URL should be: https://api-beta.us.emtelligent.com/mcp?api-key=<your-api-key>

Open a new chat window and enter the following example prompt which should trigger a tool call to the NLP API MCP server. You will need to enable the tool in the “Search and Tools” menu and allow access to the MCP server on a pop-up window.

Give me the SNOMED codes for the attached medical report.
<clinical_note>
Gen: elderly male in NAD. Oriented x3.
HEENT: NCAT. Sclera anicteric. PERRL. MMM, OP with petechiae on posterior palate and 1X0.5cm submucosal hardened/firm lesion on anterior roof of mouth with 2mm ulcer over center. Missing teeth.
</clinical_note>

Available Tools

Get SNOMED ids

Get SNOMED-CT concept identifiers that are mentioned in a given clinical note.

NAME:
get_snomed_ids

PARAMETERS:
report (str) – Any clinical note where you want to get SNOMED concept ids like from a Radiology report, SOAP note, Discharge Summary or Pathology report

RETURNS:
Table of snomed concepts and mentions.

RETURN TYPE:
str

Get ICD-10 ids

NAME:
get_icd10_ids

PARAMETERS:
report (str) – Any clinical note where you want to get ICD-10 codes or concept ids like from a Radiology report, SOAP note, Discharge Summary or Pathology report

RETURNS:
Table of ICD-10 codes and mentions and additional mapping advice to identify more specific ICD-10 codes based on the document context.

RETURN TYPE:
str

Use Cases

  1. ICD-10 coding.
  2. SNOMED-CT coding.

Natural Language Interface

No special syntax required — just ask the AI assistant to find SNOMED or ICD-10 codes in a clinical report. You can also combine coding requests to take advantage of the specific output provided by the emtelligent Medical Language Engine that allows more precise coding output. An example of a prompt that uses both of the tools is:

Give me the ICD-10 codes for the attached medical report; Use the SNOMED codes to resolve the IF conditions in the ICD-10 output:

Privacy and Compliance

  • None of your data is retained by the tool.
  • None of your data is used to improve the tool in any way including retraining or data retention.

Support

Please email support@emtelligent.com for technical support and info@emtelligent.com for business inquiries.

Contact us!

We would love your feedback. Please email us if you need help with your use cases and application of our tools to healthcare data.