Secure Hybrid Medical Agent System with MCP
Architect and implement a LangGraph-orchestrated hybrid AI-human agent system designed for initial patient intake, preliminary symptom analysis, and diagnostic support within a virtual care environment. A core requirement is secure integration with mock Electronic Health Records (EHR) and up-to-date medical knowledge bases, achieved via a Multi-Cloud Protocol (MCP) enabled tool. The system must intelligently route complex cases for human physician review, showcasing adaptive reasoning and robust human-AI collaboration.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a graph-based multi-agent system using LangGraph with Claude Opus 4.1 for secure, hybrid AI-human patient interaction and diagnostic support, integrated via MCP for EHR access.
How work is evaluated
Evaluation will assess the system's accuracy in processing patient cases, the security and correctness of EHR integration, the intelligence of diagnostic support, and the effectiveness of human-AI handoff.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Master LangGraph for building stateful, cyclic agent workflows tailored for complex medical processes.
Develop extended thinking pipelines using Claude Opus 4.1 for advanced medical reasoning and differential diagnosis.
Design and orchestrate seamless human-AI collaboration checkpoints within diagnostic pathways.
Implement MCP-enabled tool integration for secure and compliant access to mock EHR systems.
Reference links and supporting material
Sample data for 1 tasks
Challenge setup
Evaluation will assess the system's accuracy in processing patient cases, the security and correctness of EHR integration, the intelligence of diagnostic support, and the effectiveness of human-AI handoff.
Challenge input
{'patient_id': 'P123', 'symptoms': 'severe headache, nausea, sensitivity to light', 'mock_ehr': {'P123': {'history': 'migraines, no known allergies...
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{'summary': '...', 'preliminary_assessment': '...', 'suggested_diagnostics': ['...'], 'questions_for_physician': ['...'], 'ehr_access_log': [{'acti...
- Checks if the output is a valid JSON adhering to the specified format.
- Verifies that EHR access is logged correctly and simulated MCP calls are acknowledged.
- Symptom Synthesis Accuracy target: 9
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- Checks if the output is a valid JSON adhering to the specified format.
- Verifies that EHR access is logged correctly and simulated MCP calls are acknowledged.
Metrics
- Symptom_Synthesis_Accuracy · Target 9
- Diagnostic_Relevance · Target 8
- Physician_Question_Quality · Target 8
- Human_Handoff_Clarity · Target 8
Workflow tasks
- Task 1
Input: {'patient_id': 'P123', 'symptoms': 'severe headache, nausea, sensitivity to light', 'mock_ehr': {'P123': {'history': 'migraines, no known allergies', 'medications': ['sumatriptan']}}}
Output: {'summary': '...', 'preliminary_assessment': '...', 'suggested_diagnostics': ['...'], 'questions_for_physician': ['...'], 'ehr_access_log': [{'action': 'read', 'data': '...', 'status': 'success'}]}
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.