How it Works Features Clinical Skills Security MCP Apps Voice Agents CMA CLI GitHub
Install
Open Source · HIPAA Compliant · Built in Go

Backbone for Healthcare AI Agents

Works with EPIC, Cerner, and any FHIR R4 system.

Works with EPIC, Cerner, and any FHIR R4 system.

AI Agents shipped faster. Developers skip the FHIR plumbing. Clinicians build Agents.

Agents shipped faster. Developers skip FHIR plumbing. Clinicians build Agents.

⭐ New: LangCare now supports Claude Managed Agents (CMA), Claude MCP Apps, Voice AI Agents and CLI

Install Now GitHub Repo
Clinical Agent (Claude, Gemini, ChatGPT or Custom)● Connected to EPIC
See it in action
40+
Clinical Skills
150+
FHIR R4 Resources
100%
Open Source
Capabilities

Enterprise Healthcare AI made simple

One MCP server. Any FHIR backend. Every AI agent. Connect Claude, ChatGPT, Gemini, or your custom agent to clinical data in minutes.

🔌

Universal FHIR Support

Connect to EPIC, Cerner, GCP Healthcare API, or any FHIR R4 server with automatic OAuth2 and SMART on FHIR authentication.

🛡️

HIPAA Compliant

PHI scrubbing, TLS 1.3, OAuth 2.0, audit logging, and zero persistent storage. Stateless proxy architecture by design.

⚙️

Generic FHIR Tools

Read, search, create, and update any of the 60+ FHIR R4 resource types including Patient, Observation, and MedicationRequest.

📖

Clinical Skills

Evidence-based workflow guides for medication management, lab interpretation, clinical decision support, and specialty care.

‹/›

Built in Go

Enterprise-grade performance. Works with Claude, ChatGPT, Gemini, and any AI agent. Supports stdio and HTTP/SSE transports.

📦

Open Source

MIT licensed and community-driven. Fully extensible for custom workflows, new FHIR providers, and integrations.

Built For

From Developers to Clinicians

Whether you're building healthcare AI products, deploying clinical agents, or creating your own workflows.

👩‍💻

Healthcare AI Developers

Build clinical applications with FHIR data access out of the box. Use the 4 generic MCP tools to read, search, create, and update any FHIR resource.

🩺

Clinicians & Doctors

Build your own agentic workflows in Claude, ChatGPT, or Gemini — no coding required. Use 40+ ready-made clinical skills.

🏥

Enterprise Healthcare Infra

Deploy a HIPAA-compliant MCP server on your infrastructure. Supports Kubernetes, Docker, and managed hosting with full audit logging.

Quick Start

Connected to your EMR in minutes

Three steps to connect any AI agent to your FHIR-based EMR system.

📥
01

Install & Run

Global install via npm and start the server.

terminal
$ npm install -g @langcare/langcare-mcp-fhir
✔ Installed successfully
$ langcare-mcp-fhir –config config.yaml
⚡ LangCare MCP FHIR Server running
→ FHIR Backend: EPIC (connected)
→ Tools: fhir_read, fhir_search, fhir_create, fhir_update
⚙️
02

Configure Your EMR

Point to your FHIR backend with a simple YAML.

config.epic.yaml
fhir_server:
  provider: "epic"
  base_url: {your-epic-fhir-base-url}
  epic:
    client_id: "XXXXXXXXXXX"
    private_key_path: {your-key-path}
    token_url: {your-epic-fhir-token-url}
✨
03

Connect to Your Agent

Add to your AI agent's MCP config and restart.

"mcpServers": {
  {
    "langcare-mcp-fhir": {
      "command": "langcare-mcp-fhir",
      "args": ["-config",
        "/path/to/config.yaml"]
    }
  }
}
Clinical Skills Library

40+ Evidence-Based Clinical Workflows

Agent-agnostic workflow guides built on USPSTF, ADA, ACC/AHA, CDC, and other society guidelines.

💊
5 skills

Medication Management

Med reconciliation, drug interactions, adherence tracking, Beers Criteria, opioid risk

CYP450MPR/PDCMME
🔬
5 skills

Lab & Diagnostics

Lab interpretation, critical values, pre-op labs, diabetes panels, renal function

ADAKDIGOCAP/CLIA
🧠
5 skills

Clinical Decision Support

Sepsis scoring, cardiovascular risk, VTE assessment, fall risk, pneumonia severity

qSOFAASCVDWells
📋
5 skills

Documentation

SOAP notes, H&P, progress notes, discharge summaries, procedure documentation

TemplatesICD-10CPT
👥
5 skills

Population Health

Panel management, quality measures, chronic disease registries, preventive care

HEDISUSPSTFCDC
🔄
5 skills

Care Coordination

Discharge planning, referrals, care gaps, transitions of care, follow-up tasks

LACEI-PASSGaps
📊
5 skills

Patient Data & Summary

Demographics, clinical summaries, problem list audits, allergy reviews, insurance

CCDCCDAReconcile
❤️
5 skills

Specialty Workflows

Prenatal care, pediatric growth, mental health screening, oncology staging, chronic pain

ACOGPHQ-9TNM
Explore All 40+ Skills
Security

Two-layer security for HIPAA Compliance

Stateless proxy architecture with enterprise-grade security at every layer.

┌─────────────┐ ┌──────────────┐ ┌─────────────┐ │ AI Agent │ │ MCP Server │ │ FHIR API │ │ Claude/GPT │────────▶│ (Go) │────────▶│ (EMR) │ └─────────────┘ └──────────────┘ └─────────────┘
Auth1: MCP Client Auth (Bearer Token / API Key)
Auth2: FHIR Backend Auth (OAuth2 / SMART on FHIR)
Supported: Bearer Token · OAuth2 · SMART on FHIR · Basic Auth · mTLS · Custom
🔒

TLS 1.3 Encryption

End-to-end encryption for all HTTP transport communications

📝

Audit Logging

HIPAA-compliant audit trail for every FHIR operation with PHI scrubbing

🔑

OAuth2 + SMART on FHIR

Native support for EPIC, Cerner, and standard OAuth2

💾

Zero Persistent Storage

Stateless proxy — no patient data is ever stored

MCP Apps

Interactive UIs inside AI Chat

Rich clinical dashboards that render directly inside MCP-capable hosts like Claude Desktop. No LLM round-trips — apps call FHIR tools directly and deterministically.

Patient Chart Review dashboard
patient_chart_review

Patient Chart Review

Full clinical dashboard with demographics, active conditions, medications, allergies, vitals trend charts (BP + weight), labs, and recent encounters — all in one view.

FHIR Explorer
fhir_explorer

FHIR Explorer

Interactive FHIR resource browser. Search, read, create, and update any of the 60+ FHIR R4 resource types with JSON detail views and inline editing.

📊

Rich Visualization

SVG charts, color-coded cards, expandable detail panels

🎛️

Interactive Controls

Search fields, date pickers, click-to-expand rows

⚡

Deterministic Fetching

Apps call FHIR tools directly — no LLM in the data loop

📦

Zero Dependencies

Single HTML file, embedded in the Go binary

🔌

Works Offline

No CDN, no external scripts — just FHIR API calls

🧩

Extensible

Build your own apps with React + TypeScript, compiled with Vite

MCP Apps are the first official extension to the Model Context Protocol — pushing MCP beyond text into true interactive experiences. The interface comes to the conversation, not the other way around.

Voice Agent

Real-time Voice AI for Patient Care

Patients talk to their health records over phone or browser. Built on PipeCat, powered by Claude, with FHIR R4 access via LangCare MCP.

Healthcare Voice Agent Architecture — PipeCat + LangCare MCP

Five layers, three auth boundaries. Speech flows through DeepGram STT → Claude → Cartesia TTS while clinical data is fetched from your EMR via LangCare MCP — all in real time.

🎙️

Browser, Phone, or Mobile

WebRTC, PSTN dial-in, or native app — patients connect however they prefer

🔊

PipeCat Voice Pipeline

DeepGram STT → Claude LLM → Cartesia TTS — orchestrated in real time

🧠

Claude + MCP Tool Calls

Claude reasons over the request and calls FHIR tools for labs, meds, conditions

🏥

LangCare MCP → EMR

FHIR R4 operations over Streamable HTTP to EPIC, Cerner, or any FHIR server

🔐

3-State Authentication

Pre-authenticated, DOB verification, or unknown caller — automatic patient matching

💬

Natural Conversations

"What were my last lab results?" — patients ask in plain language, get spoken answers

⚡

Real-time FHIR Access

Labs, meds, conditions, allergies, vitals — fetched and spoken in seconds

🔄

Swap Any Layer

DeepGram, Google, Azure STT — Claude, Gemini, OpenAI — Cartesia, ElevenLabs TTS

☁️

Deploy Anywhere

PipeCat Cloud, GCP Cloud Run, GKE, or any container platform

📞

Enterprise Ready

Genesys AudioHook, PSTN dial-in, contact center integration

Claude Managed Agents

LangCare Skills as Claude Managed Agents

LangCare's 40 clinical skills run natively as CMA custom skills — progressive disclosure, versioning, and composability on Anthropic's managed agent platform. 9 lean agents, each referencing 5 skills, connected to your EMR via the LangCare MCP Server. Read More →

Claude Managed Agents — LangCare Skills-First Architecture
ConceptDescription
AgentThe model, system prompt, tools, MCP servers, and skills. 9 lean agents — one per clinical category + 1 general triage agent.
EnvironmentA configured container template with packages, network access to the LangCare MCP Server, and mounted files.
SessionA running agent instance within an environment, performing a specific clinical task and generating outputs.
Skills40 CMA custom skills mapped from LangCare's Clinical Skills Library — progressive disclosure, versioned, workspace-wide.
MCP ConnectorConnects agents to the LangCare MCP Server for fhir_search, fhir_read, fhir_create, and fhir_update operations.
🧠

Skills-First Design

40 clinical skills with progressive disclosure — metadata always loaded, instructions on demand

🏗️

Managed Infrastructure

No agent loop to build — Anthropic handles tool execution, sandboxing, and runtime

🔗

MCP Connector

Agents connect to LangCare MCP for real-time FHIR R4 access to your EMR

🤖

Multi-Agent

Orchestrate specialist agents — medication, labs, CDS — in a single session thread

📊

9 Clinical Agents

One agent per category: meds, labs, CDS, docs, population health, care coord, specialty

🔒

Enterprise Security

Vaults for credentials, isolated containers, HIPAA-compliant audit trail

CLI

LangCare CLI — Any Agent Framework

Wraps 4 FHIR operations as CLI subcommands over HTTP, handling the MCP session handshake internally. Built for AI agents that don't speak MCP natively — LangChain, smolagents, CrewAI, AutoGen, and any framework that can call a subprocess.

MCP-Native — connect directly (no CLI needed) direct /mcp
Claude Desktop PipeCat Claude Code
Speak MCP natively → connect to /mcp endpoint directly.
Auto-discovers fhir_read · fhir_search · fhir_create · fhir_update
Non-MCP — use CLI as subprocess tool subprocess
OpenClaw smolagents CrewAI AutoGen any framework
Call langcare fhir search ... as subprocess → clean JSON on stdout.
CLI handles MCP handshake — agents never see JSON-RPC or SSE.
Architecture — how it flows
● MCP-Native agents
Claude Desktop ──┐
PipeCat ─────────┤
                    │
                    ▼
● Non-MCP agents
LangChain ──────┐
smolagents ─────┤
CrewAI ─────────┤
AutoGen ────────┤
                   ▼
CLI subprocess
langcare fhir search ...
HTTP POST /mcp
(initialize → tools/call)
LangCare MCP Server
local or Fly.io
/mcp endpoint
↓ FHIR R4 REST
EMR
Epic · Cerner · GCP FHIR
Skills work identically:
Skill says "use fhir_search
to find active meds"

MCP-native: calls /mcp directly

Non-MCP: calls
langcare fhir search
MedicationRequest
--query "patient=123
&status=active"

→ identical FHIR result
Installation
pip install langcare-cli
Option 1 — pip from GitHub
# Recommended — isolated environment
$ pipx install "langcare-cli @ git+https://github.com/
  langcare/langcare-mcp-fhir.git#subdirectory=cli"
$ langcare --version
✔ langcare-cli 0.1.0
Option 2 — from source
$ cd cli/
$ python -m venv .venv
$ source .venv/bin/activate
$ pip install -e .
✔ langcare fhir --help
4 FHIR Commands
fhir search
→ fhir_search
langcare fhir search <Type>
resourceType
--query <params>
langcare fhir search Patient --query "name=John&birthdate=gt1990"
langcare fhir search Observation --query "patient=123&category=laboratory"
fhir read
→ fhir_read
langcare fhir read <Type> <id>
resourceType
<id> resource ID
langcare fhir read Patient 123
langcare fhir read Observation abc-456
fhir create
→ fhir_create
langcare fhir create <Type>
resourceType
--data <json|@file>
langcare fhir create Observation --data '{"resourceType":"Observation"...}'
langcare fhir create Patient --data @patient.json
fhir update
→ fhir_update
langcare fhir update <Type> <id>
resourceType <id>
--data <json|@file>
langcare fhir update Patient 123 --data '{"id":"123"...}'
langcare fhir update MedicationRequest abc --data @med.json
View CLI on GitHub
Pricing

Open source first, Enterprise ready

Start free with the full-featured open source server. Scale with enterprise support when you need it.

Most Popular

Open Source

Free
Self-hosted and fully featured
Get Started →
  • ✓Complete FHIR integration
  • ✓40+ clinical skills library
  • ✓HIPAA-compliant architecture
  • ✓MIT license
  • ✓Community support
  • ✓Unlimited usage

On-Prem / Private Cloud

Custom
AWS, GCP or Azure
Contact Us →
  • ✓Everything in Open Source
  • ✓On-premises deployment
  • ✓Private cloud installation
  • ✓Custom FHIR providers
  • ✓Architecture consulting
  • ✓HIPAA compliance review
  • ✓Dedicated Slack channel
  • ✓SLA agreements

LangCare Cloud

Custom
Fully managed
Contact Sales →
  • ✓Everything in On-Prem
  • ✓Dedicated infrastructure
  • ✓24/7 monitoring & alerting
  • ✓Automatic scaling
  • ✓Backup & disaster recovery
  • ✓SOC 2 compliance
  • ✓99.9% uptime SLA

Transform healthcare with
AI-powered clinical workflows

Join healthcare organizations worldwide using LangCare to connect AI agents to FHIR-based EMRs.

Install Now View on GitHub