LangCare MCP FHIR: agent access to medical records
LangCare MCP FHIR
A Go MCP server whose four FHIR R4 operations give an agent read and write access to Epic, Cerner, OpenEMR and GCP Healthcare API
- 71 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent reads and edits medical charts, and a wrong record or a leak of patient data has legal consequences
- The fhir_create and fhir_update tools write to the clinic's live system with the server account's rights
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/langcare-mcp-fhirDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent reads and edits medical charts, and a wrong record or a leak of patient data has legal consequences
- The fhir_create and fhir_update tools write to the clinic's live system with the server account's rights
This is third-party code. Review the repository files before installing.
What it does
LangCare MCP FHIR sits between an agent and the FHIR server of a medical information system. It is written in Go and exposes four tools: fhir_read, fhir_search, fhir_create and fhir_update. They work with any FHIR R4 resource: Patient, Observation, Medication and more than 60 other types. The server handles EMR sign-in and token refresh itself: Bearer, OAuth2, SMART on FHIR Backend Services with private_key_jwt, Basic. It has no database of its own and proxies to your FHIR server. Transports are stdio and Streamable HTTP. In logs the server scrubs patient identifiers, keeps an audit trail and supports mTLS and per-client rate limits. Two interactive MCP apps for Claude Desktop are embedded in the binary: fhir_explorer for browsing resources and patient_chart_review with demographics, conditions, medications, labs and blood pressure and weight charts. The repository also carries extras: more than 40 clinical skills as workflow guides, nine agents for Claude Managed Agents, a Pipecat voice agent and a langcare command-line client for frameworks without MCP. The server works without them.
Who it is for. Teams building AI assistants for clinics and connecting them to Epic, Cerner, OpenEMR or Google Cloud Healthcare API.
Good fit when
- You need to find a patient by name and birth date and show their latest labs with reference ranges
- A clinician wants a note draft built from chart data, which the agent turns into FHIR resources
- A clinical assistant prototype runs on a test OpenEMR or HAPI server and you want to try real FHIR
- A team wants to show a clinician a patient chart with graphs right in the Claude Desktop window
- A LangChain or CrewAI agent reaches FHIR through the langcare command-line client without native MCP
Not a fit when
- You need access to a system that does not serve FHIR R4, since the server speaks only that protocol
- The agent must not change the chart: create and update operations are available by default, so limit them with the EMR account permissions
- You want treatment advice without a clinician: the clinical skills give query structure, and the decision stays with a specialist
- Patient data may not go to your model provider under a regulator's rules
Example request
Find the patient Smith born in 1985 and show their latest blood tests with reference ranges.MCP
- Transport
- stdio, http
- Authentication
- API key
Limitations
The server needs a YAML file for the chosen system and credentials on the EMR side, and it will not start without them. The agent's rights equal that account's rights, so create a separate role for read-only mode. The last commit in the repository was made in April 2026. The clinical skills serve as agent guides and do not count as a medical opinion. The MCP client passes its own check with a Bearer token or API key, independent of the server's sign-in to the EMR.
How to disable. Remove the server from the Claude Desktop configuration and restart the app. The global package comes off with npm uninstall -g @langcare/langcare-mcp-fhir.
FAQ
How do I connect LangCare MCP FHIR to Claude Desktop?
Install the package with npm install -g @langcare/langcare-mcp-fhir, prepare a YAML file from the template for your system and add a server to claude_desktop_config.json with the command langcare-mcp-fhir and the arguments -config and the file path. After the app restarts the four FHIR tools appear.
Which systems does the server work with?
Epic, Cerner, OpenEMR, Google Cloud Healthcare API and any FHIR R4 server. The configs folder in the repository has an example settings file for each.
Does the server store patient data?
No, it works as a proxy without its own database, and personal data in logs is scrubbed by default. The records themselves stay in your EMR.
Do I need the clinical skills from the repository?
No, they are optional and the server works fully without them. The skills describe workflows such as medication reconciliation or qSOFA scoring, and their text goes into the agent's system prompt.
What is the langcare command-line client for?
It serves agents that do not speak MCP: LangChain, smolagents, CrewAI, AutoGen. The client turns the four FHIR operations into subcommands and returns clean JSON, taking the server address and token from LANGCARE_SERVER and LANGCARE_API_KEY.
Discussion
No questions or reviews yet.
Related
MCP serversFHIR MCP tools for an agent: build, read and validate resources from a patient bundle with an actionable validation report.
Anthropic's official plugin for healthcare work: ICD-10-CM coding, prior authorization, FHIR access and PubMed search
Medical MCP
medical-mcp
MCP server for medical sources: FDA and EMA drug data, WHO statistics, PubMed and clinical trials without API keys.
PubMed MCP searches biomedical papers, retrieves available full text and formats citations with retraction and correction notices.
A medical code reference for your agent: search, map and navigate SNOMED CT, ICD-10, RxNorm and LOINC concepts across vocabularies
A Go MCP server and REST API for databases: PostgreSQL, MySQL, ClickHouse, Snowflake, BigQuery and more, with personal data masking and a cache
README in short
The README covers npm installation, settings templates for four kinds of systems, the four FHIR tools and a two-layer sign-in security model. Separate sections describe the embedded MCP apps and the catalog of clinical skills.