FHIR MCP Server by CData

by cdatasoftware

Not rated
GitHub

About

A read-only MCP server for FHIR, enabling LLMs to query live FHIR data. Requires the CData JDBC Driver for FHIR.

Details

Author
cdatasoftware
Categories
Database, Other, API

Setup

Install FHIR MCP Server by CData in your MCP client (Claude Desktop, Cursor, Windsurf, and others).

Repository: https://github.com/cdatasoftware/fhir-mcp-server-by-cdata

Follow the installation instructions in the repository README, then restart your MCP client.

CData's Model Context Protocol (MCP) Server for FHIR

:heavy_exclamation_mark: This project builds a local, read-only MCP server. For fully remote MCP access, check out the first managed MCP platform:CData Connect AI. For locally hosted, full read, write, update, delete, and action capabilities and a simplified setup, check out ourCData MCP Server for FHIR.

We created this read-only MCP Server to allow LLMs (like Claude Desktop) to query live data FHIR supported by theCData JDBC Driver for FHIR.

CData JDBC Driver connects to FHIR by exposing them as relational SQL models.

This server wraps that driver and makes FHIR data available through a simple MCP interface, so LLMs can retrieve live information by asking natural language questions — no SQL required.
- Clone the repository:

git clone https://github.com/cdatasoftware/fhir-mcp-server-by-cdata.git cd fhir-mcp-server-by-cdata

- Navigate to thelibfolder in the installation directory, typically:

- (Windows)C:\Program Files\CData\CData JDBC Driver for FHIR\
- (Mac/Linux)/Applications/CData JDBC Driver for FHIR/

-

Run the commandjava -jar cdata.jdbc.fhir.jarto open the Connection String utility.

Configure the connection string and click "Test Connection"

Note:If the data sources uses OAuth, you will need to authenticate in your browser.

Once successful, copy the connection string for use later.

- Prefix- a prefix to be used for the tools exposed
- ServerName- a name for your server
- ServerVersion- a version for your server
- DriverPath- the full path to the JAR file for your JDBC driver
- DriverClass- the name of the JDBC Driver Class (e.g. cdata.jdbc.fhir.FHIRDriver)
- JdbcUrl- the JDBC connection string to use with the CData JDBC Driver to connect to your data (copied from above)
- Tables- leave blank to access all data, otherwise you can explicitly declare the tables you wish to create access for

Prefix=fhir ServerName=CDataFHIR ServerVersion=1.0 DriverPath=PATH\TO\cdata.jdbc.fhir.jar DriverClass=cdata.jdbc.fhir.FHIRDriver JdbcUrl=jdbc:fhir:InitiateOAuth=GETANDREFRESH; Tables=

-

Create the config file for Claude Desktop ( claude_desktop_config.json) to add the new MCP server, using the format below. If the file already exists, add the entry to themcpServersin the config file.

{ "mcpServers": { "{classname_dash}": { "command": "PATH\\TO\\java.exe", "args": [ "-jar", "PATH\\TO\\CDataMCP-jar-with-dependencies.jar", "PATH\\TO\\fhir.prp" ] }, ... } }
{ "mcpServers": { "{classname_dash}": { "command": "/PATH/TO/java", "args": [ "-jar", "/PATH/TO/CDataMCP-jar-with-dependencies.jar", "/PATH/TO/fhir.prp" ] }, ... } }

If needed, copy the config file to the appropriate directory (Claude Desktop as the example).Windows

cp C:\PATH\TO\claude_desktop_config.json %APPDATA%\Claude\claude_desktop_config.json
cp /PATH/TO/claude_desktop_config.json /Users/{user}/Library/Application\ Support/Claude/claude_desktop_config.json'

Run or refresh your client (Claude Desktop).

Note:You may need to fully exit or quit your Claude Desktop client and re-open it for the MCP Servers to appear.
- Run the follow the command to run the MCP Server on its own

java -jar /PATH/TO/CDataMCP-jar-with-dependencies.jar /PATH/TO/Salesforce.prp

Note:The server usesstdioso can only be used with clients that run on the same machine as the server.

Once the MCP Server is configured, the AI client will be able to use the built-in tools to read, write, update, and delete the underlying data. In general, you do not need to call the tools explicitly. Simply ask the client to answer questions about the underlying data system. For example:

- "What is the correlation between my closed won opportunities and the account industry?"
- "How many open tickets do I have in the SUPPORT project?"
- "Can you tell me what calendar events I have today?"

The list of tools available and their descriptions follow:

In the definitions below,{servername}refers to the name of the MCP Server in the config file (e.g.{classname_dash}above).

- {servername}_get_tables- Retrieves a list of tables available in the data source. Use the{servername}_get_columnstool to list available columns on a table. The output of the tool will be returned in CSV format, with the first line containing column headers.
- {servername}_get_columns- Retrieves a list of columns for a table. Use the{servername}_get_tablestool to get a list of available tables. The output of the tool will be returned in CSV format, with the first line containing column headers.
- {servername}_run_query- Execute a SQL SELECT query

If you are scripting out the requests sent to the MCP Server instead of using an AI Client (e.g. Claude), then you can refer to the JSON payload examples below – following the JSON-RPC 2.0 specification - when calling the available tools.

{ "jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": { "name": "source_get_tables", "arguments": {} } }
{ "jsonrpc": "2.0", "id": 2, "method": "tools/call", "params": { "name": "source_get_columns", "arguments": { "table": "Account" } } }
{ "jsonrpc": "2.0", "id": 3, "method": "tools/call", "params": { "name": "source_run_query", "arguments": { "sql": "SELECT * FROM [Account] WHERE [IsDeleted] = true" } } }

- If you cannot see your CData MCP Server in Claude Desktop, be sure that you have fully quit Claude Desktop (Windows: use the Task Manager, Mac: use the Activity Monitor)
- If Claude Desktop is unable to retrieve data, be sure that you have configured your connection properly. Use the Connection String builder to create the connection string (see above) and copy the connection string into the property (.prp) file.
- If you are having trouble connecting to your data source, contact theCData Support Team.
- If you are having trouble using the MCP server, or have any other feedback, join the
CData Community.

This MCP server is licensed under the MIT License. This means you are free to use, modify, and distribute the software, subject to the terms and conditions of the MIT License. For more details, please see theLICENSEfile in the project repository.

Search and query patient Electronic Health Record (EHR) data using SMART on FHIR.

Access medical knowledge-bases and drug information from eka.care. Requires API credentials.

Enhanced Medication Information MCP Server

Provides real-time access to FDA drug data, including shortages, labeling, and recalls, via the openFDA API.

Open-source, conformance-aware MCP server for FHIR R4 and R5.

NIH clinical trials and FDA adverse event reports. 4 MCP tools for health research.

Interact with AWS Amplify Gen2 data models using natural language and Cognito authentication.

A Python-based MCP server that connects to an Elasticsearch index with Apple HealthKit data.

Query Apple Health data using natural language and SQL.

Access live U.S. congressional data from the Congress.gov API.

A nutrition analysis platform integrating Canada's Food Guide recipes with Health Canada's official nutrition databases.

No reviews yet — be the first

Sign in to leave a review

Use Google, GitHub, or an email account so ratings stay tied to real people.

Email sign in

No reviews posted yet.