40 AI Agents — Live

The Future of
AI-Powered Medicine
is Already Here

Agentic Hospital is a fully autonomous AI healthcare system — a root coordinator intelligently triages patients and routes them to 40 specialist AI agents, including triage, pharmacy, and discharge planning agents, each equipped with domain-specific clinical tools.

0 Specialist Agents
0 Clinical Tools
0 Patient Profiles
24/7 Always Available
Triage Result
Acuity Level 2 — Urgent
Routing to Cardiology…
hospital_coordinator — adk session
Patient: "Chest pain, left arm numbness, sweating"
Running triage_assessment()…
ACUITY 2 → cardiology_agent
Cardiology Agent active…
Tools: assess_cardiac_risk() • get_lab_results()
HEART score: 6 — High Risk
Differential: STEMI, NSTEMI, unstable angina
✓ Emergency care advised
ALL 40 AGENTS ONLINE
Web Search
ACC/AHA 2024 Guidelines loaded
browser-use • gemini-2.0-flash
0
AI Agents
0+
Clinical Tools
0
ESI Triage Levels
24/7
Continuous Operation

Specialized Agents for Hospital Operations

Beyond clinical departments — intelligent agents for triage, pharmacy, discharge, and bed management

🏥
Triage Nurse Agent
ESI v4 Triage System
Performs Emergency Severity Index (ESI) v4 triage scoring — the gold standard for emergency department prioritization. Calculates acuity levels 1-5, assigns waiting priority, and manages the patient queue.
calculate_esi_score() record_nurse_triage() assign_waiting_priority()
💊
Pharmacy Agent
Medication Safety
Clinical pharmacist agent for medication verification, dispensing, and reconciliation. Validates against formulary, checks allergies, screens for drug interactions, and generates TTA prescriptions.
verify_medication_order() dispense_medication() medication_reconciliation() generate_tta_prescription()
📋
Discharge Planning
Care Continuation
Manages the complete discharge workflow — assesses readiness, generates discharge summaries, sends GP letters, and arranges community services (PT, OT, district nursing).
check_discharge_criteria() generate_discharge_summary() send_gp_letter() arrange_community_services()

Seven steps from symptom to care plan

Every interaction follows an evidence-based 7-step clinical process — from intake to follow-up — mirroring real hospital protocols.

STEP 01
📋
Symptom Intake
Patient describes symptoms in natural language. The coordinator extracts chief complaint, duration, severity, and associated symptoms using structured NLP.
STEP 02
AI Triage
5-level ESI acuity scoring detects red flags, life threats, and urgency. Immediate 911 education is triggered for emergencies — no delays.
STEP 03
🔀
Smart Routing
The coordinator agent selects the optimal specialist agent from 40 agents (35 departments + triage + pharmacy + discharge), performs context-preserving handoff with full patient state.
STEP 04
🔬
Clinical Assessment
Specialist agent retrieves patient records, runs specialty tools (ECG analysis, GFR calculation, etc.), and builds differential diagnoses with confidence levels.
STEP 05
🌐
Guideline Search
browser-use agent queries latest ACC/AHA, NICE, WHO guidelines in real time — ensuring recommendations are always current and evidence-based.
STEP 06
📄
SOAP Note Generation
Structured clinical summary: Subjective, Objective, Assessment, Plan — including differentials, must-not-miss conditions, workup, and red flags.
STEP 07
📅
Care Coordination
Drug interaction checks, dose calculations, appointment scheduling, and specialist referrals — all automated within a single conversational turn.
CONTINUOUS
🧠
Adaptive Learning
Each agent refines reasoning through clinical feedback loops, guideline updates, and cross-department consultation for complex multi-system cases.

Every major medical specialty,
AI-powered and ready

Each department agent has its own persona, specialty tools, evidence-based protocols, and clinical reasoning framework.

Clinical-grade AI, built for scale

Every system component is designed around real medical workflows, safety-first principles, and production-grade reliability.

See what doctors see — from ECGs to CT scans

The vision-capable image analysis tool processes 13 medical image types with type-specific clinical prompts, delivering structured findings like an experienced radiologist.

ABCDE criteria for skin lesion evaluation with melanoma risk scoring
ECG interpretation — rhythm, axis, interval, ST changes, ischemia patterns
Radiological reading — X-ray, CT, MRI, PET/CT, bone scan, ultrasound
Pathology slides — IHC markers, histopathology patterns, biopsy interpretation
🫀
ECG Analysis
ST elevation leads II, III, aVF — Inferior STEMI. HEART score: 8. Recommend cath lab activation.
🔬
Pathology
Adenocarcinoma pattern, moderately differentiated. TTF-1 positive. EGFR mutation testing recommended.
🧠
Brain MRI
Diffusion restriction right MCA territory. Acute ischemic stroke. mRS baseline assessment initiated.
🩻
Chest X-Ray
Right lower lobe consolidation. Air bronchograms present. CAP Severity: PSI Class III.
# Drug Interaction Check
medications = ["Warfarin", "Aspirin", "Clopidogrel"]
# Result
⚠ MAJOR INTERACTION DETECTED
Warfarin + Clopidogrel: Bleeding Risk ↑↑↑
Aspirin + Warfarin: Triple antithrombotic risk
→ Hematology consult recommended
→ INR monitoring every 3 days
# Dose Calculation
Vancomycin • 68kg • CrCl: 42 mL/min
Dose: 1000mg q24h (renal adjusted)
Target trough: 10–15 μg/mL

Clinical decision support that protects patients

Every agent enforces multi-layer safety checks — drug interactions, renal dosing, emergency escalation — with transparent, explainable reasoning at every step.

Drug interaction checking across all patient medications with severity grading
Renal & hepatic dose adjustment with weight-based pediatric dosing
Validated scoring systems — HEART, Wells, SOFA, NIHSS, PHQ-9, GCS and more
Cognitive bias detection — anchoring, premature closure, framing effect checks

Built on cutting-edge AI infrastructure

Purpose-built with the most advanced AI frameworks available — engineered for reliability, speed, and clinical accuracy.

🤖
Google ADK
Google's Agent Development Kit provides the multi-agent orchestration framework — enabling hierarchical agent trees, context-preserving handoffs, and tool execution.
google-adk multi-agent
Gemini 2.0 Flash
All 40 specialist agents and the root coordinator run on Gemini 2.0 Flash — delivering 1M token context, multimodal vision, and lightning-fast inference for real-time clinical use.
gemini-2.0-flash multimodal
🌐
browser-use
AI-powered browser automation enables real-time web search for latest clinical guidelines, drug approvals, and research publications — running safely in a daemon thread.
browser-use playwright
🔗
LiteLLM + OpenRouter
Universal LLM gateway enabling seamless model switching between Gemini, GPT-4o, Claude, and local Ollama models — zero code changes required for model upgrades.
litellm openrouter
🐍
Python Tool Ecosystem
101+ synchronous clinical tool functions covering vitals recording, lab retrieval, SOAP note generation, medication dosing, triage scoring, pharmacy verification, and discharge planning.
python-dotenv langchain
🗄️
Clinical Data Layer
10 detailed mock patient profiles (P001–P010) with complete labs, medications, allergies, and histories. In-memory drug interaction database for instant lookups.
in-memory DB mock EHR

End‑to‑end agentic hospital blueprint

A layered architecture where a single root coordinator orchestrates specialist agents, clinical tools, and data services — from patient interfaces to EHR systems.

Patient Interfaces
💬 Web & mobile chat
🗂️ EHR in-basket messages
📞 Voice / IVR front-door
🔌 API & webhook clients
All channels normalized into a single patient session and routed to the coordinator.
Ingress & Safety
✅ AuthN / AuthZ
🛡️ PHI / PII redaction
📊 Audit & trace logging
⏱️ Rate limiting & abuse filters
Root Coordinator
hospital_coordinator
Gemini 2.0 Flash · Orchestrator
🏥
Controls the 4-phase flow: Registration → Triage → Specialist → Pharmacy / Discharge. Maintains patient state and routing decisions.
Agent Pools
Clinical specialist agents
🩺
Cardiology, Neurology, Oncology, ICU, Emergency, Pediatrics, Endocrine, Nephrology, Rheumatology, Hematology, ID, Derm, Ophthalmology, ENT and more.
cardiology_agent icu_agent oncology_agent id_agent neurology_agent
Operational agents
⚙️
Triage, Pharmacy, Discharge, Bed Management, Admission, Scheduling, Documentation.
triage_nurse_agent pharmacy_agent discharge_agent admission_agent
Tools & Data Layer
Clinical tool functions (101+)
get_patient_info(), record_vitals(), get_lab_results(), analyze_medical_image(), generate_soap_note(), check_drug_interactions(), dose_calculator(), triage_scores(), order_investigations() and more.
Data & external services
  • Mock EHR store · P001–P010 profiles
  • Medication & interaction database
  • Guideline fetcher via browser-use
  • LLM gateway via LiteLLM / OpenRouter
Downstream systems
EHR writeback, task & inbox creation, SMS / email notifications, analytics, observability, governance dashboards.
4‑phase patient journey
Phase 1 · Registration intake_agent
Phase 2 · Triage triage_nurse_agent
Phase 3 · Specialist *_{specialty}_agent
Phase 4 · Pharmacy & Discharge pharmacy_agent · discharge_agent
Cross‑cutting safety & governance
Guardrails run across every phase: input validation, guideline‑aligned recommendations, explanation traces, hallucination checks, cognitive bias checks, human‑in‑the‑loop overrides, and full audit logs for every agent decision.

Grounded in peer-reviewed science

Agentic Hospital is built on a synthesis of the latest academic research in agentic AI and healthcare AI systems.

Memorial Sloan Kettering · 2025
Next-Generation Agentic AI for Transforming Healthcare
Karunanayake et al. Comprehensive framework for deploying agentic AI in clinical settings with safety guardrails and human-in-the-loop design.
Tsinghua University · 2025
Agent Hospital: A Simulacrum of Hospital with Evolvable Medical Agents
Li et al. Multi-agent hospital simulation demonstrating how AI doctors and nurses can collaborate for improved diagnostic accuracy and patient outcomes.
IEEE Access · 2026
Agentic AI in Healthcare & Medicine: A Seven-Dimensional Taxonomy
Vatsal et al. Taxonomic framework covering perception, reasoning, planning, memory, tool use, communication, and safety for healthcare AI agents.
⚠️ Educational & Research Use Only. Agentic Hospital is an AI system for educational, research, and informational purposes. It does not provide medical advice, diagnosis, or treatment. Always consult a licensed healthcare professional for medical decisions. In emergencies, call 911 or your local emergency services immediately.

Deploy an AI hospital
for your patients today

From triage to discharge — give every patient a guided, structured AI experience backed by 40 specialist agents and real-time clinical guidelines.