MIT License

技术解读

以下内容由 GitRead 中文社区 编辑团队基于项目 README 与官方资料整理, 结合中文开发者的常见使用场景补充技术选型建议。具体以项目官方文档为准。

多智能体协作不是堆模型,而是拼编排

当大多数团队还在用单个大模型跑 RAG、写摘要时,微软在 GitHub 上开源了一个名为 Multi-Agent Custom Automation Engine Solution Accelerator 的项目(简称 MACAE),试图解决一个更棘手的问题:如何让多个 AI Agent 围绕一个复杂目标高效协作,而不是各自为战。

这个仓库的价值不在代码量,而在于它提供了一个可落地的多 Agent 编排范式——一个拿来就能跑、能改、能扩展的完整骨架。


项目背景与痛点:多 Agent 系统为什么难做?

单 Agent 应用的处理能力有上限:上下文窗口受限、工具调用链过长时容易丢失焦点、无法并行处理子任务。即便强行塞进多个 Agent,新问题接踵而至——

  • Agent 之间如何通信? 如果每个 Agent 各干各的,输出结果靠什么汇聚?谁来决定分支走向?
  • 状态一致性怎么保证? 多个 Agent 并发修改同一组数据,出现冲突由谁仲裁?
  • 任务如何分解与调度? 一个"帮我安排下周行程"的请求,需要哪些 Agent 介入?串行还是并行?

MACAE 给出的答案是:用基础设施来约束协作,而不是用自然语言来磨合。


核心技术揭秘:编排引擎 + 全链路可观测的基础架构

该项目基于微软自家的 Agent Framework(继承自 AutoGen),并深度绑定 Azure Foundry、Azure Cosmos DB 以及基础设施服务。这不是随意的技术选型,每一个都有明确目标:

  1. Agent Framework(编排核心):提供 Agent 的注册、调度和事件响应机制,支持多种对话模式(单轮、多轮、群聊式),是任务分发的大脑。
  2. Azure Cosmos DB(状态中枢):这是最巧妙的设计。MACAE 没有沿用 AutoGen 的内存式会话保留,而是将任务状态、Agent 中间输出、会话上下文持久化到 Cosmos DB。这意味着系统可以水平伸缩、可以故障恢复,甚至可以在多个实例之间共享任务视图。
  3. Azure Foundry(可观测与质量保障):通过 Foundry 的评估和监控能力,实现对 Agent 行为的跟踪和响应质量的度量。相当于给多 Agent 系统装了监控仪表盘。

数据流大致如下:

graph TD A[用户输入请求] --> B{"意图识别与任务分解"} B -->|"需要子任务"| C[编排引擎分配 Agent 组] B -->|"单任务直答"| G[单一 Agent 执行] C --> DB[("Cosmos DB 会话与<br/>任务状态持久化")] DB --> D[子 Agent 执行具体逻辑] D --> E{"结果是否需要聚合?"} E -->|"是"| F[聚合器合并输出] E -->|"否"| H[直接返回] F --> I[Azure Foundry 评估与<br/>行为日志记录] G --> I I --> J[最终响应给用户]

核心设计思想可以从三个维度提炼:

  • 状态外置而非内存驻留。 Cosmos DB 让 Agent 之间的交接不再是函数调用的延续,而是基于共享状态的数据流转。这让超长任务、跨实例接管的可能性成为现实。
  • 事件驱动而非线性调用。 Agent Framework 的群聊机制天然支持"某个 Agent 完成后自动唤醒另一个"的异步协作模式,而不是简单的 A 调 B。
  • 可评估性优先。 生产级 AI 应用最怕黑盒。Foundry 的集成确保每个 Agent 的行为可审查、可回放、可量化。

功能亮点与差异:和 LangChain/LangGraph 这类框架的区别

LangGraph 确实解决了编排的 DAG(有向无环图)问题,但它的状态管理默认基于内存,且框架本身不提供端到端的监控方案。MACAE 的差异化在于:

  • 全链路基础设施化。从对话管理、状态存储到性能监控,全部由 Azure 云服务承接。你不需要自己搭 Redis 或开发日志收集器,架构层面天然为生产做好准备。
  • 群聊式协作原生支持。Agent Framework 的群聊范式(GroupChat)让多个 Agent 之间可以互相"讨论",比固定 pipeline 的链式调用更接近人类的协作方式,也更容易处理开放域任务。
  • 解决方案加速器定位。它的意义不是给你一堆可调用的库函数,而是一个包含推荐架构、示例 Agent 技能、交互模式的模板项目。你复制、修改、部署,即可获得一个完整可用的多 Agent 任务执行系统,而不是从零编织技术栈。

应用场景与落地建议

最适合接入这类系统的场景具备以下特征:任务结构开放、需要多步骤操作、数据必须持久化。典型如企业级 IT 运维助手(自动排查故障、分配工单)、内容生产流水线(资料收集→初稿生成→合规审查→发布)、数据分析驾驶舱(根据自然语言指令拉取多源数据并生成报告)。

落地时的三个提醒:

  1. 别把所有计算过程都甩给 Agent。能够用确定性代码完成的校验、格式化逻辑,不要塞进模型。编排引擎只做决策判断,不做繁琐的数据搬运。
  2. 设计好会话的生存周期。Cosmos DB 上的状态数据需要定义保留和清理策略,否则高并发场景下成本会快速膨胀。
  3. 延迟要预估。多 Agent 协作的端到端时延显著高于单模型调用。如果业务场景对响应速度敏感,建议增加"直答分支",只对复杂请求才进入群聊模式。

综合评价

优点突出:架构清晰、云原生适配度高、状态管理设计老练,特别适合已经押注 Azure 生态的中大型团队。它省去的是从 MVP 到生产环境的漫长基础设施打磨期。

缺点同样明显——强绑定微软云。如果你的技术栈不在 Azure 上,这套方案能直接复用的部分非常有限。此外,Agent Framework 的社区规模相比 LangChain 仍有差距,遇到边缘场景时的排障资料会相对稀缺。最后,它的抽象层级较高,对底层推理逻辑的理解成本不小,新手直接上手容易被绕晕。

一句话总结:这是给懂云架构的人准备的 Agent 生产参考书,不是给初学者上的 Prompt 入门课。

关于

GitHub 原始描述(英文)

The Multi-Agent Custom Automation Engine Solution Accelerator is an AI-driven system that manages a group of AI agents to accomplish tasks based on user input. Powered by Microsoft Agent Framework, Azure Foundry, Azure Cosmos DB, and infrastructure services, it provides a reference application, allowing you to hit the ground running.

描述取自 GitHub 官方 API,保持仓库作者书写的原文。

简要分析

热度分
2630
价值分
794
活跃状态
活跃

语言占比

Bicep
CSS
Dockerfile
HTML
JavaScript
PowerShell
Python
Shell
TypeScript

Release

README

Multi-Agent Custom Automation Engine Solution Accelerator

Welcome to the Multi-Agent Custom Automation Engine solution accelerator, designed to help businesses leverage AI agents for automating complex organizational tasks. This accelerator provides a foundation for building AI-driven orchestration systems that can coordinate multiple specialized agents to accomplish various business processes.

When dealing with complex organizational tasks, users often face significant challenges, including coordinating across multiple departments, maintaining consistency in processes, and ensuring efficient resource utilization.

The Multi-Agent Custom Automation Engine solution accelerator allows users to specify tasks and have them automatically processed by a group of AI agents, each specialized in different aspects of the business. This automation not only saves time but also ensures accuracy and consistency in task execution.



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.

Solution overview

The solution leverages Azure OpenAI Service, Azure Container Apps, Azure Cosmos DB, and Azure Container Registry to create an intelligent automation pipeline. It uses a multi-agent approach where specialized AI agents work together to plan, execute, and validate tasks based on user input.

Solution architecture

image

Agentic architecture

image

Additional resources

Agent Framework Documentation

Azure AI Foundry Documentation

Azure Container App documentation


Key features

Click to learn more about the key features this solution enables
  • Allows people to focus on what matters
    By doing the heavy lifting involved with coordinating activities across an organization, people’s time is freed up to focus on their specializations.

  • Enabling GenAI to scale
    By not needing to build one application after another, organizations are able to reduce the friction of adopting GenAI across their entire organization. One capability can unlock almost unlimited use cases.

  • Applicable to most industries
    These are common challenges that most organizations face, across most industries.

  • Efficient task automation
    Streamlining the process of analyzing, planning, and executing complex tasks reduces time and effort required to complete organizational processes.



Quick deploy

How to install or deploy

Follow the quick deploy steps on the deployment guide to deploy this solution to your own Azure subscription.

Note: This solution accelerator requires Azure Developer CLI (azd) version 1.18.0 or higher. Please ensure you have the latest version installed before proceeding with deployment. Download azd here.

Note: If deploying from a local environment (Option D in the deployment guide), this solution accelerator requires Bicep CLI version 0.33.0 or higher to compile infrastructure templates. Install Bicep.

Click here to launch the deployment guide

Open in GitHub Codespaces Open in Dev Containers Open in Visual Studio Code Web

Note: Some tenants may have additional security restrictions that run periodically and could impact the application (e.g., blocking public network access). If you experience issues or the application stops working, check if these restrictions are the cause. In such cases, consider deploying the WAF-supported version to ensure compliance. To configure, Click here.

⚠️ Important: Check Azure OpenAI Quota Availability
To ensure sufficient quota is available in your subscription, please follow quota check instructions guide before you deploy the solution.


Prerequisites and Costs

To deploy this solution accelerator, ensure you have access to an Azure subscription with the necessary permissions to create resource groups and resources. Follow the steps in Azure Account Set Up.

Check the Azure Products by Region page and select a region where the following services are available: Microsoft Foundry Service, Foundry Models, Azure AI Search, and Search Semantic Ranker.

Here are some example regions where the services are available: Australia East, East US2, France Central, Japan East, Norway East, Sweden Central, UK South, West US.

Pricing varies per region and usage, so it isn’t possible to predict exact costs for your usage. The majority of the Azure resources used in this infrastructure are on usage-based pricing tiers. However, Azure Container Registry has a fixed cost per registry per day.

Use the Azure pricing calculator to calculate the cost of this solution in your subscription. Review a sample pricing sheet for the architecture.

Product Description Cost
Microsoft Foundry Powers the AI agents for task automation Pricing
Azure AI Search Powers vector based data search Pricing
Azure Container Apps Hosts the web application frontend Pricing
Azure Cosmos DB Stores metadata and processing results Pricing
Azure Container Registry Stores container images for deployment Pricing

⚠️ Important: To avoid unnecessary costs, remember to take down your app if it’s no longer in use, either by deleting the resource group in the Portal or running azd down.



Business Scenario

image

Companies maintaining and modernizing their business processes often face challenges in coordinating complex tasks across multiple departments. They may have various processes that need to be automated and coordinated efficiently. Some of the challenges they face include:

  • Difficulty coordinating activities across different departments
  • Time-consuming process to manually manage complex workflows
  • High risk of errors from manual coordination, which can lead to process inefficiencies
  • Lack of available resources to handle increasing automation demands

By using the Multi-Agent Custom Automation Engine solution accelerator, users can automate these processes, ensuring that all tasks are accurately coordinated and executed efficiently.

Business value

Click to learn more about what value this solution provides
  • Process Efficiency
    Automate the coordination of complex tasks, significantly reducing processing time and effort.

  • Error Reduction
    Multi-agent validation ensures accurate task execution and maintains process integrity.

  • Resource Optimization
    Better utilization of human resources by focusing on specialized tasks.

  • Cost Efficiency
    Reduces manual coordination efforts and improves overall process efficiency.

  • Scalability
    Enables organizations to handle increasing automation demands without proportional resource increases.

Use Case

Click to learn more about what use cases this solution provides
Use Case Persona Challenges Summary/Approach
Product Marketing Marketing Executive Marketing release plans require input from multiple teams (engineering, design, compliance), which often leads to delays and misalignment. Traditional planning involves repetitive tasks like drafting timelines, assigning owners, and validating compliance, which are prone to mistakes. Through an agentic approach the Multi-agent speeds up release planning by automating repetitive tasks and enhances collaboration with dynamic agent teams that adapt to campaign needs.
Onboarding Employee HR Manager Traditional onboarding involves multiple disconnected steps—HR paperwork, IT setup, compliance training—which often require manual coordination and lead to delays. Manual steps slow collaboration and increase overhead, reducing efficiency and employee experience. Designed to streamline the complex process of bringing new hires into an organization using a modular, agentic architecture.
Retail Remediation Customer Success Manager This approach is seeking a smarter way to manage workflows and ensure customer satisfaction. Challenges include difficulty engaging the right agents, inefficiencies from manual processes, fragmented data, and limited security controls. Using the Multi-Agent Custom Automation Engine, the Customer Success Manager recruits intelligent agents to analyze satisfaction and recommend steps for remediation. The platform’s modular design and reasoning capabilities allow for analyzing dependencies, planning transitions, and reducing manual rework.
RFP Reviewer VP of Finance Reviewing RFPs manually under tight deadlines is challenging. Compliance checks and risk assessments slow progress, leaving critical gaps and delaying decisions. Multi-Agent RFP Reviewer is an intelligent platform addresses these issues by deploying AI agents to review RFPs, identify risks, recommend remediation, and execute actions seamlessly - transforming a labor-intensive process into a fast, accurate, and compliant workflow.
Contract Compliance Reviewer Compliance Counsel Reviewing Contracts for compliance manually under tight deadlines is challenging. Compliance checks and risk assessments slow progress, leaving critical gaps and delaying decisions. Multi-Agent Contract Compliance Reviewer is an intelligent platform addresses these issues by deploying AI agents to review Contracts, identify risks, recommend remediation, and execute actions seamlessly - transforming a labor-intensive process into a fast, accurate, and compliant workflow.
Content Generation Marketing Creative Accelerate your marketing content creation by leveraging AI to interpret creative briefs and generate on-brand, multimodal content. The solution helps marketing teams reduce time-to-market for campaigns by automating the creation of compliant marketing copy and images grounded in product data. This solution leverages a team of agents to interpret creative briefs, retrieve product information, generate marketing content, and validate brand compliance.



Supporting documentation

Security guidelines

This template uses Azure Key Vault to store all connections to communicate between resources.

This template also uses Managed Identity for local development and deployment.

To ensure continued best practices in your own repository, we recommend that anyone creating solutions based on our templates ensure that the GitHub secret scanning setting is enabled.

You may want to consider additional security measures, such as:


Cross references

Check out similar solution accelerators

Solution Accelerator Description
Document Knowledge Mining Extract structured information from unstructured documents using AI
Modernize your Code Automate the translation of SQL queries between different dialects
Conversation Knowledge Mining Enable organizations to derive insights from volumes of conversational data using generative AI

💡 Want to get familiar with Microsoft’s AI and Data Engineering best practices? Check out our playbooks to learn more

Playbook Description
AI playbook The Artificial Intelligence (AI) Playbook provides enterprise software engineers with solutions, capabilities, and code developed to solve real-world AI problems.
Data playbook The data playbook provides enterprise software engineers with solutions which contain code developed to solve real-world problems. Everything in the playbook is developed with, and validated by, some of Microsoft’s largest and most influential customers and partners.

Provide feedback

Have questions, find a bug, or want to request a feature? Submit a new issue on this repo and we’ll connect.


Responsible AI Transparency FAQ

Please refer to Transparency FAQ for responsible AI transparency details of this solution accelerator.


Disclaimers

This release is an artificial intelligence (AI) system that generates text based on user input. The text generated by this system may include ungrounded content, meaning that it is not verified by any reliable source or based on any factual data. The data included in this release is synthetic, meaning that it is artificially created by the system and may contain factual errors or inconsistencies. Users of this release are responsible for determining the accuracy, validity, and suitability of any content generated by the system for their intended purposes. Users should not rely on the system output as a source of truth or as a substitute for human judgment or expertise.

This release only supports English language input and output. Users should not attempt to use the system with any other language or format. The system output may not be compatible with any translation tools or services, and may lose its meaning or coherence if translated.

This release does not reflect the opinions, views, or values of Microsoft Corporation or any of its affiliates, subsidiaries, or partners. The system output is solely based on the system’s own logic and algorithms, and does not represent any endorsement, recommendation, or advice from Microsoft or any other entity. Microsoft disclaims any liability or responsibility for any damages, losses, or harms arising from the use of this release or its output by any user or third party.

This release does not provide any financial advice, legal advice and is not designed to replace the role of qualified client advisors in appropriately advising clients. Users should not use the system output for any financial decisions, legal guidance or transactions, and should consult with a professional financial advisor and or legal advisor as appropriate before taking any action based on the system output. Microsoft is not a financial institution or a fiduciary, and does not offer any financial products or services through this release or its output.

This release is intended as a proof of concept only, and is not a finished or polished product. It is not intended for commercial use or distribution, and is subject to change or discontinuation without notice. Any planned deployment of this release or its output should include comprehensive testing and evaluation to ensure it is fit for purpose and meets the user’s requirements and expectations. Microsoft does not guarantee the quality, performance, reliability, or availability of this release or its output, and does not provide any warranty or support for it.

This Software requires the use of third-party components which are governed by separate proprietary or open-source licenses as identified below, and you must comply with the terms of each applicable license in order to use the Software. You acknowledge and agree that this license does not grant you a license or other right to use any such third-party proprietary or open-source components.

To the extent that the Software includes components or code used in or derived from Microsoft products or services, including without limitation Microsoft Azure Services (collectively, “Microsoft Products and Services”), you must also comply with the Product Terms applicable to such Microsoft Products and Services. You acknowledge and agree that the license governing the Software does not grant you a license or other right to use Microsoft Products and Services. Nothing in the license or this ReadMe file will serve to supersede, amend, terminate or modify any terms in the Product Terms for any Microsoft Products and Services.

You must also comply with all domestic and international export laws and regulations that apply to the Software, which include restrictions on destinations, end users, and end use. For further information on export restrictions, visit https://aka.ms/exporting.

You acknowledge that the Software and Microsoft Products and Services (1) are not designed, intended or made available as a medical device(s), and (2) are not designed or intended to be a substitute for professional medical advice, diagnosis, treatment, or judgment and should not be used to replace or as a substitute for professional medical advice, diagnosis, treatment, or judgment. Customer is solely responsible for displaying and/or obtaining appropriate consents, warnings, disclaimers, and acknowledgements to end users of Customer’s implementation of the Online Services.

You acknowledge the Software is not subject to SOC 1 and SOC 2 compliance audits. No Microsoft technology, nor any of its component technologies, including the Software, is intended or made available as a substitute for the professional advice, opinion, or judgment of a certified financial services professional. Do not use the Software to replace, substitute, or provide professional financial advice or judgment.

BY ACCESSING OR USING THE SOFTWARE, YOU ACKNOWLEDGE THAT THE SOFTWARE IS NOT DESIGNED OR INTENDED TO SUPPORT ANY USE IN WHICH A SERVICE INTERRUPTION, DEFECT, ERROR, OR OTHER FAILURE OF THE SOFTWARE COULD RESULT IN THE DEATH OR SERIOUS BODILY INJURY OF ANY PERSON OR IN PHYSICAL OR ENVIRONMENTAL DAMAGE (COLLECTIVELY, “HIGH-RISK USE”), AND THAT YOU WILL ENSURE THAT, IN THE EVENT OF ANY INTERRUPTION, DEFECT, ERROR, OR OTHER FAILURE OF THE SOFTWARE, THE SAFETY OF PEOPLE, PROPERTY, AND THE ENVIRONMENT ARE NOT REDUCED BELOW A LEVEL THAT IS REASONABLY, APPROPRIATE, AND LEGAL, WHETHER IN GENERAL OR IN A SPECIFIC INDUSTRY. BY ACCESSING THE SOFTWARE, YOU FURTHER ACKNOWLEDGE THAT YOUR HIGH-RISK USE OF THE SOFTWARE IS AT YOUR OWN RISK.

评论

暂无评论