OMOP MCP Server
Map clinical terminology to standardized OMOP CDM concepts using LLMs and the OMOPHub vocabulary API.
Add to your client
Copy the config for your MCP client and paste it into its config file.
git clone https://github.com/OHNLP/omop_mcp.gitPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"omop-mcp-server": {
"command": "uv",
"args": [
"--directory",
"<path-to-local-repo>",
"run",
"omop_mcp"
]
}
}
}Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- uv installed (https://docs.astral.sh/uv/getting-started/installation/)
- Clone of the OHNLP/omop_mcp repository
- OMOPHUB_API_KEY from omophub.com for vocabulary lookups
- An LLM provider API key (e.g. OpenAI, Anthropic, Azure OpenAI, OpenRouter, Groq, Hugging Face, or Google), configured via a local .env file
About OMOP MCP Server
An agentic MCP server that maps free-text clinical terms to standardized OMOP CDM concepts. It combines the OMOPHub vocabulary API with an LLM to search, suggest, and select the best standard concept, with automatic domain-specific vocabulary prioritization and batch CSV mapping. Distributed by the OHNLP organization and run locally via uv as a stdio server for Claude Desktop.
Tools & capabilities (3)
find_omop_conceptSearch the OMOP vocabulary directly to look up standardized concepts.
map_clinical_terminologyIntelligently map a free-text clinical term to a standardized OMOP concept, optionally guided by the target OMOP table/field, returning the concept id, code, name, class, validity, domain, vocabulary, and a rationale.
batch_mappingMap multiple clinical concepts at once from a CSV file.
When to use it
- Standardizing free-text clinical terms (conditions, measurements, drugs) into OMOP CDM concept IDs
- Looking up OMOP vocabulary concepts during ETL of source data to the OMOP Common Data Model
- Batch-mapping a CSV of source terms to standard concepts for a data harmonization project
- Assisting OHDSI/OMOP researchers and data engineers with concept selection and vocabulary prioritization
Security notes
Requires an OMOPHUB_API_KEY (from omophub.com) and at least one LLM provider key (OpenAI, Anthropic, Azure OpenAI, OpenRouter, Groq, Hugging Face, or Google). Credentials are supplied via a local .env file copied from .env.template; do not commit the .env file. Clinical terms you map are sent to the configured LLM provider and the OMOPHub API.
OMOP MCP Server FAQ
What vocabulary backend does it use?
It uses the OMOPHub API (omophub.com) for vocabulary search, concept suggestion, and terminology mapping. You need an OMOPHUB_API_KEY.
Which LLM providers are supported?
The .env template includes keys for OpenAI, Anthropic, Azure OpenAI, OpenRouter, Groq, Hugging Face, and Google; fill in only the provider(s) you intend to use.
How do I run it in Claude Desktop?
Clone the repo, install uv, copy .env.template to .env and fill in your keys, then add the provided mcpServers block to claude_desktop_config.json with the path to your local repo.
Can I try it without installing anything?
Yes, a hosted demo (OMapper) is available at https://omapper.vercel.app/.
Alternatives to OMOP MCP Server
Compare all alternatives →
Official MCP server providing persistent, file-backed knowledge-graph memory across sessions.
Structured step-by-step reasoning tool for breaking problems into revisable thought sequences.
Fully managed remote server for AWS docs, blogs, What's-New and Well-Architected guidance — no key.
Compare OMOP MCP Server with: