采用Microsoft代理框架，采用四个Foundry托管代理（合规、临床、覆盖、综合）进行付款方AI辅助的事先授权审查。
概览
一款多智能体AI解决方案，为健康计划支付方实现自动化。四个专业——合规、临床审查、覆盖和综合——评估PA请求与保险政策的匹配，并在2分钟内生成可审计的批准/待处理建议。由、和构建。 基于Anthropic事先授权审查技能的决策政策和评估方法：宽松模式决策政策、按标准进行MET/NOT_MET/不足评估、置信度评分、渐进门槛评估、结构化审计轨迹、NCCI捆绑风险标记、服务类型分类以及提供者专业程序适宜性作为可审计标准。 使用这些模板创建的任何AI解决方案，您都有责任评估所有相关风险，并遵守所有适用法律和安全标准。更多信息请参阅代理服务和代理框架的透明度文件。
概述（AI 翻译）
一款多代理AI解决方案，自动化健康计划支付方的事先授权。四个专业的Foundry托管代理——合规、临床审查、覆盖和综合——评估事前授权请求与保险政策，并在2分钟内生成可审计的批准/待处理建议。由Microsoft Foundry、Microsoft代理框架（MAF）、Azure容器应用和MCP医疗数据服务器构建。
该解决方案支持两种运行时模式：
| 模式 | 如何开始 | 会发生什么 |
|---|---|---|
| Foundry 托管代理（推荐） | azd up |
代理注册于 Microsoft Foundry 托管代理（更新预览版）;每个代理都有一个专用端点，后端通过 AIProjectClient.get_openai_client(agent_name=...) 调用该端点，编号为 DefaultAzureCredential。Foundry 管理容器生命周期。 |
本地 / Docker Composedocker compose up |
所有4个代理容器 + 后端 + 前端都本地运行——无需Azure部署。 |
本页由 AI 翻译，技术细节请以原文为准。
README
以下为原文 README 节选，完整内容请查看原文。
Prior Authorization Multi-Agent Solution Accelerator
A multi-agent AI solution that automates prior authorization for health plan payers. Four specialized Foundry Hosted Agents — Compliance, Clinical Reviewer, Coverage, and Synthesis — evaluate PA requests against coverage policies and produce auditable approve/pend recommendations in under 2 minutes. Built with Microsoft Foundry, the Microsoft Agent Framework (MAF), Azure Container Apps, and MCP healthcare data servers.
The solution supports two runtime modes:
| Mode | How to start | What happens |
|---|---|---|
| Foundry Hosted Agent (recommended) | azd up |
Agents are registered with Microsoft Foundry Hosted Agents (refreshed preview); each agent gets a dedicated endpoint that the backend invokes through AIProjectClient.get_openai_client(agent_name=...) with DefaultAzureCredential. Foundry manages the container lifecycle. |
| Local / Docker Compose | docker compose up |
All 4 agent containers + backend + frontend run locally — no Azure deployment needed. |
Decision policy and evaluation methodology adapted from the Anthropic prior-auth-review-skill: LENIENT mode decision policy, per-criterion MET/NOT_MET/INSUFFICIENT evaluation, confidence scoring, progressive gate evaluation, structured audit trails, NCCI bundling risk flagging, service-type classification, and provider specialty-procedure appropriateness as an auditable criterion.
[!NOTE] With any AI solutions you create using these templates, you are responsible for assessing all associated risks and for complying with all applicable laws and safety standards. Learn more in the transparency documents for Agent Service and Agent Framework.
Features
- Multi-agent parallel execution — Four specialized agents complete a full PA review in under 2 minutes; Compliance and Clinical agents run concurrently via
asyncio.gather - Foundry Hosted Agents — Each specialist agent is independently containerized and deployed on Microsoft Foundry; Foundry manages the container lifecycle
安装
This server does not publish a one-line install command.
Open the repository installation guide 新手？先看 mcpskill.org 教程 →