Healthcare Agents

51 AI agents that actually know healthcare administration.
Real CFR citations. Real CMS payment models. Real deliverables. Not a chatbot with a healthcare skin.

Drop 22,000+ lines of healthcare administration expertise into any AI coding tool. These agents handle HIPAA compliance audits, medical coding reviews, revenue cycle management, CMS cost reports, payer contract analysis, and 45 other healthcare admin specialties -- with the depth of an MHA graduate, not a generic prompt.

Built on the conventions of agency-agents by Michael Sitarzewski.

Install in 30 Seconds

One command. Auto-detects your tools.

curl -fsSL https://raw.githubusercontent.com/ajhcs/healthcare-agents/main/install.sh | bash

Or target a specific tool:

git clone https://github.com/ajhcs/healthcare-agents.git
cp healthcare-agents/agents/*.md ~/.claude/agents/    # Claude Code
cp healthcare-agents/agents/*.md .cursor/rules/        # Cursor
cp healthcare-agents/agents/*.md .github/instructions/  # Copilot

Then just ask:

"Activate the Compliance Officer and audit our HIPAA Security Rule compliance against 45 CFR 164.308"

Full setup instructions for all 12+ tools: INSTALL.md

What Makes These Different

22,000+ lines of dense domain knowledge across 51 agents. Average 420 lines/agent of real expertise -- regulatory citations, payment model math, system-specific workflows -- not filler.

Real Output, Not Summaries

Ask the 340B Program Manager about contract pharmacy compliance and you get this:

340B CONTRACT PHARMACY COMPLIANCE ASSESSMENT
=============================================

1. ENTITY ELIGIBILITY VERIFICATION
   - Covered entity type: [Federally Qualified Health Center / DSH Hospital / ...]
   - 340B ID: [HRSA-assigned ID]
   - Registration status in 340B OPAIS: [Active / Pending]

2. CONTRACT PHARMACY NETWORK REVIEW
   - Total contract pharmacies registered: [count]
   - Chain vs. independent mix: [ratio]
   - HRSA contract pharmacy registration dates: [verify all current]

3. COMPLIANCE RED FLAGS
   [ ] Duplicate discounts -- Medicaid claims cross-referenced against 340B purchases
   [ ] Diversion -- prescriptions for non-eligible patients filled at contract pharmacy
   [ ] GPO prohibition violations (for DSH hospitals under 100 beds)
   [ ] Missing contract pharmacy agreements or expired registrations

4. RECOMMENDED ACTIONS
   ...

Every agent produces structured, actionable deliverables -- compliance assessments, financial models, audit checklists, gap reports. Not summaries of what a deliverable would look like.

Use Cases

HIPAA compliance audit -- The Compliance Officer walks through 45 CFR 164.308 administrative safeguards, 164.310 physical safeguards, and 164.312 technical safeguards. Produces a gap assessment with specific remediation steps and regulatory citations.

Revenue cycle denial management -- The Revenue Cycle Specialist analyzes denial patterns using CARC/RARC codes, identifies root causes (registration errors, medical necessity, timely filing), and builds an appeals workflow with payer-specific requirements.

Medical coding accuracy review -- The Medical Coding Specialist audits documentation for CC/MCC capture, E/M level support under 2021+ guidelines, and NCCI edit compliance. Flags undercoding and overcoding with CPT/ICD-10 specifics.

CMS cost report preparation -- The Healthcare Finance Manager structures CMS-2552 worksheets, calculates cost-to-charge ratios, and identifies reclassification opportunities for wage index and DSH adjustments.

Payer contract negotiation -- The Contract Analyst models reimbursement scenarios across fee-for-service, capitation, and value-based arrangements. Compares rates against Medicare benchmarks and MGMA data.

Eval Status

10 of 51 agents currently score 80+ on a frozen eval rubric. Each agent goes through iterative question-answer-judge-improve cycles: the scorer generates domain-specific exam questions, grades the agent's answers, identifies weak areas, and produces a targeted improvement brief. An editor model patches only the weak spots while preserving the agent's identity and voice.

Full scores and iteration history: eval/results.tsv

Infrastructure: The /eval skill runs the loop. A frozen rubric at eval/rubric.md keeps scoring consistent across runs. A calibration pipeline ensures scorer reliability before any agent edits land.

The 51 Agents

Self-Improving Agents

These agents are not static. The repo ships a self-improvement kit for Claude Code and Codex that runs automated eval-and-improve loops:

bash healthcare-agents/scripts/install-self-improvement-kit.sh /path/to/your/project

Then run from either tool:

• Claude Code: /eval revenue-medical-coding-specialist
• Codex: Run the healthcare self-improvement loop for revenue-medical-coding-specialist

The loop generates domain-specific exam questions, scores the agent's answers against a frozen rubric, identifies weak areas, and patches the prompt -- while preserving the agent's identity and voice. In runtimes that support native subagents, the strongest model scores while a faster model edits.

This does not require the Python eval harness. See INSTALL.md for the full setup.

Compatible Tools

Validation

bash scripts/lint-agents.sh

FAQ

Can AI agents handle HIPAA compliance?

These agents cite specific CFR sections (45 CFR 160, 162, 164), know the difference between addressable and required implementation specifications, and produce structured gap assessments. They do not handle PHI or make legal determinations -- they give you the regulatory framework and audit methodology that a compliance officer would use.

What healthcare certifications and regulations do these agents cover?

The agents cover CMS Conditions of Participation, Joint Commission standards, NCQA HEDIS measures, MIPS/QPP quality programs, CMS Five-Star ratings, OASIS/PDGM for home health, MDS/PDPM for skilled nursing, HRSA 340B requirements, OIG compliance program guidance, and dozens of other regulatory frameworks. Each agent's specialty section lists its specific coverage.

Do these agents work with Epic, Cerner, or other EHR systems?

The Epic Applications Analyst has deep knowledge of Epic build/configuration, Bridges integration, and Caboodle/Cogito analytics. Other agents reference EHR-specific workflows where relevant (CPOE for medication safety, CDI queries for documentation improvement, eCQMs for quality reporting). The agents are not EHR plugins -- they provide the healthcare administration expertise that complements your EHR workflows.

How are agents evaluated and improved?

Each agent goes through iterative eval cycles: a scorer generates domain-specific exam questions (regulatory scenarios, calculation problems, workflow decisions), grades the agent's responses on a frozen rubric, and produces an improvement brief. An editor model patches weak areas while preserving the agent's identity. 10 agents currently score 80+ out of 100. Scores and methodology are fully transparent in eval/results.tsv.

Can I use these agents for revenue cycle management?

Six agents cover the full revenue cycle: Medical Coding Specialist (ICD-10, CPT, DRG, HCC coding), Revenue Cycle Specialist (denials management, A/R optimization, CARC/RARC analysis), Healthcare Finance Manager (CMS-2552 cost reports, budgeting), Contract Analyst (payer reimbursement modeling), 340B Program Manager (drug discount compliance), and Chargemaster Analyst (CDM maintenance, price transparency).

Contributing

New agents, deeper regulatory citations, updated CMS rules -- all welcome. See CONTRIBUTING.md.

Disclaimer

These agents provide healthcare administration knowledge for informational and operational purposes. They do not provide clinical advice, legal opinions, or handle PHI. Regulations change -- verify against primary sources (CMS.gov, Federal Register, state regulatory bodies).

License

Apache 2.0 -- see LICENSE.