内容来源：https://blog.n8n.io/deploy-n8n-agents-that-show-up-as-members-of-the-team-inside-microsoft-apps/

内容总结：

n8n与微软Agent 365正式集成：AI代理化身“虚拟同事”入驻Microsoft 365办公套件

随着微软Agent 365功能的全面上线，n8n平台用户现在可以构建直接嵌入Microsoft 365应用的AI代理。这些代理能够像真实团队成员一样参与日常工作——你可以在Microsoft Teams中@提及它们，在Outlook中添加它们为邮件收件人，或在Word文档中标记它们以协同工作。

技术亮点：企业级安全管控下的灵活代理设计

微软Agent 365为每个代理分配独立的Entra ID身份，使其能像同事一样被授予SharePoint和Teams频道的访问权限，同时通过Purview与Defender系统处理身份验证、访问策略及合规性管理。管理员可在Microsoft 365管理中心统一管理代理生命周期、规则策略与分析数据。n8n的可视化工作流画布则负责定义代理的核心能力：系统提示词、第三方工具连接、业务逻辑编排及具体执行动作。当用户在Teams中向代理发送消息时，n8n将跨系统协调工作流，并将结果直接回传至对话界面。

生态合作与行业价值

n8n副总裁Ben Kiziltug指出：“Microsoft 365承载着大量日常工作，Agent 365让n8n用户构建的代理不仅能在后台自动化运行，更能直接参与实际工作流程。”微软Copilot与代理生态系统副总裁Srini Raghavan表示：“n8n的加入丰富了Agent 365生态，为开发者提供了安全可控的方式，在协作、沟通和创作场景中部署AI代理。”

实战案例：跨系统协同场景

以客户支持负责人为例：每天早晨需检查Zendesk夜间升级工单、从Salesforce提取客户信息、查看PagerDuty基础设施告警，最终在Teams发布摘要并邮件跟进。现在只需在Teams频道@提及n8n驱动的Agent 365代理，后者即可自动调取工单、交叉验证账户数据、检测活跃事件，最后生成聊天消息与包含处理步骤及责任人的邮件草稿。所有操作在n8n画布上通过拖拽节点完成——Zendesk、Salesforce、PagerDuty等工具只需配置一次凭证即可复用。

深度扩展：工作流复用与自定义开发

现有n8n工作流可作为子工作流被代理直接调用。例如客户数据丰富流程（含Salesforce数据提取、合同状态校验、健康评分）无需在代理中重建，仅需连接即可调用。团队还可通过MCP客户端节点接入Datadog等观测工具，让代理在故障排查时获取实时服务性能数据。当内置节点不满足需求时，用户可在工作流任意节点插入JavaScript或Python代码，实现自定义数据转换、业务逻辑或通过HTTP请求节点调用内部服务。

双重治理体系：微软层+工作流层安全管控

代理治理分为两层：微软层面负责身份管理（Entra ID）、条件访问策略、数据保护（Purview）及威胁检测（Defender），同时提供代理注册、分析与策略管理；n8n层面则聚焦工作流级管控，包括基于角色的访问控制（RBAC）、凭证管理、输入验证护栏、PII检测及异常处理流程。支持数据驻留与合规要求的团队可自托管n8n，完全控制数据存储与处理位置。所有代理执行过程均生成详细审计记录，呈现每个决策、调用工具及返回数据，全程可视、可测、可编辑。

部署指南

现有通过独立节点使用微软服务的n8n工作流仍可正常运行。使用Agent 365需在微软侧完成三步：在Entra ID注册应用、通过Agent 365 CLI创建代理蓝图、获取租户管理员审批。n8n侧只需添加Agent 365触发器节点并连接工具即可。审批通过后，代理将出现在Teams、Outlook、SharePoint和Word中，团队成员可随时@提及使用。相关文档已在n8n官网与微软Agent 365页面开放。

中文翻译：

随着Microsoft Agent 365正式上线，n8n用户现在可以构建AI智能体，这些智能体能够直接出现在Microsoft 365应用内部，在人们日常工作的场景中发挥作用。现在，你可以从Microsoft Teams中@提及在n8n构建的智能体，将其添加到Outlook的邮件中，或在Word中标记它们，从而将它们融入你的工作流程。

Microsoft Agent 365为每个智能体分配独立的Entra ID，因此员工可以像授权团队成员一样，授予其访问SharePoint和Teams频道等Microsoft 365服务的权限。微软通过Purview和Defender管理身份、访问策略及合规性。Agent 365还负责智能体的生命周期管理操作、规则与策略，以及在Microsoft 365管理中心进行分析；通过熟悉的工具，提供了一种有效且可扩展的方式来大规模管控智能体。而n8n画布则是你设计智能体行为的地方：包括它的系统提示词、与第三方工具的连接、将这些元素串联起来的逻辑，以及它要执行的具体操作。当有人在Teams中给智能体发送消息时，n8n会协调各个所需系统完成工作，并将结果直接返回到对话中。

Microsoft 365承载了大量日常工作。借助Agent 365，n8n用户可以构建不仅能在后台自动执行任务，更能直接参与到实际工作中去的智能体。而对于微软开发者来说，n8n提供了一个可视化构建器以及超过一千个原生集成，用于将智能体连接到他们的其他技术栈。

——n8n副总裁 Ben Kiziltug

此次合作建立在2025年11月Ignite大会上宣布的合作基础之上。随着Agent 365全面上市，任何拥有Microsoft 365租户和n8n的组织都可以使用其全部功能。希望在其Azure环境中进行部署的团队，可以在Microsoft Marketplace获取n8n。这意味着可以直接从n8n画布中受控访问Agent 365的MCP工具，并且能够部署显示为微软应用内团队成员的智能体。

我们很高兴欢迎n8n加入Agent 365生态系统。借助Agent 365，像n8n这样的合作伙伴可以将智能体直接带入Microsoft 365的工作流程中。它为开发者提供了一种安全、可控的方式来构建智能体，使其出现在人们已经进行协作、沟通和创作的地方。

——微软公司副总裁，Microsoft Copilot与智能体生态系统负责人 Srini Raghavan

了解如何使用Microsoft Agent 365触发器节点在n8n中构建智能体，将其连接到你的团队所依赖的工具，并让它开始在Microsoft Teams、Outlook、Word和SharePoint中工作。包含现场演示和问答环节。

你可以构建什么

设想一位客户支持负责人，她每天早晨需要检查Zendesk的夜间升级工单，从Salesforce提取客户账户信息，查看PagerDuty是否有任何可能影响客户的基础设施未解决问题，然后手动在Teams中发布摘要，并通过邮件跟进客户团队。借助由n8n驱动的Agent 365智能体，她只需在Teams频道中@提及该智能体，其余工作都由它来处理。智能体会拉取工单、交叉关联客户数据，并检查三个系统中是否有任何活跃事件，然后发布聊天消息并草拟一封包含后续步骤和负责人分配的邮件。

在n8n画布上，通过将各种服务作为Microsoft Agent 365节点的工具添加，协调这些服务变得非常直接。Zendesk通过拖放节点连接到智能体，该节点带有自己的凭证，可以一次配置并在多个工作流中重复使用。Salesforce、PagerDuty也是如此添加。智能体根据对话及其系统提示词中提供的上下文，调用所需的任何工具。像发送邮件和在Teams中发布消息这类Microsoft服务，则通过Agent 365的MCP服务器实现，并可在Microsoft Agent 365节点的配置中通过开关启用。

现有的n8n工作流和第三方MCP工具使这一切更加强大。假设那位支持负责人已经构建了一个n8n工作流，该工作流通过从Salesforce拉取数据、交叉核对合同状态以及评估客户健康评分来丰富客户记录。这个工作流可以被打包成一个子工作流，并成为智能体可调用的工具。她无需在智能体内重建该逻辑，只需连接它即可。如果团队使用Datadog进行可观测性，其MCP服务器可以通过MCP客户端节点作为工具连接，从而让智能体在处理客户问题时，能够实时了解服务性能。

可视化设计，按需编码，全面管控

当你需要超越内置节点的便利性时，n8n允许你在工作流的任何环节插入JavaScript或Python代码。利用此选项进行自定义数据转换和业务逻辑处理，或通过HTTP请求节点调用内部服务。在有意义的地方编写代码，在其他地方使用可视化节点，并在同一个工作流中自由地组合它们。

治理在两个层面上运作。Agent 365在微软层面处理身份、生命周期管理、安全性和合规性：包括Entra ID、条件访问策略、通过Purview进行数据保护，以及通过Defender进行威胁检测，还有在Microsoft 365管理中心中的智能体注册、分析和策略。n8n在工作流层面处理治理，包括工作流访问的基于角色的访问控制（RBAC）、已连接服务的凭证管理、用于验证输入和执行安全检查（如PII检测）的护栏，以及出现问题时的工作流错误处理。两个层面协同工作，使智能体从身份到执行都受到管理。如果你有数据驻留或合规性要求，n8n可以自托管在你自己的基础设施上，让你完全掌控数据的存储位置和处理方式。

对于那些难以洞察AI智能体内部运作情况的团队来说，在n8n中构建意味着你可以看到智能体在运行时的实际行为。每次执行都会产生一份详细的智能体输出和行为记录，包括它做出的每一个决策、调用的每一个工具，以及从这些工具返回的数据。整个链条都是可见的、可测试的，并且任何一步都可以编辑。

开始使用

如果你已经通过单个节点将n8n与Microsoft服务配合使用，那么这些工作流仍然可以正常工作。Agent 365为那些希望其智能体出现在微软应用内部并像团队成员一样交互的团队提供了新的途径。

让Agent 365智能体运行起来，需要在微软方面进行一些初始设置：在Entra ID中注册应用，通过Agent 365 CLI创建智能体蓝图，以及获取管理员对租户内智能体的批准。在n8n方面，添加Agent 365触发器节点并连接你的工具。原生节点、MCP连接的服务和子工作流都可以作为智能体可调用的工具。一旦智能体获得批准，它就会出现在Teams、Outlook、SharePoint和Word中，随时供你的团队@提及并投入使用。

Microsoft Agent 365触发器节点现已在n8n中可用。请参阅n8n文档来设置和构建你的智能体，并查看Microsoft的Agent 365文档以了解如何为你的组织进行配置。

英文来源：

With the general availability of Microsoft Agent 365, n8n users can now build AI agents that show up inside the Microsoft 365 applications where people get work done. Now you can @mention agents built in n8n from Microsoft Teams, add them to emails in Outlook, or tag them in Word to bring them into your work.
Microsoft Agent 365 gives each agent its own Entra ID, so employees can grant it access to Microsoft 365 services like SharePoint and Teams channels the same way they would a teammate. Microsoft handles identity, access policies, and compliance through Purview and Defender. Agent 365 also handles agent lifecycle management actions, rules and policies, analytics in the Microsoft 365 Admin Center; providing an effective and scalable way to control agents at scale through familiar tools. The n8n canvas is where you design what the agent actually does: its system prompt, its connections to third-party tools, the logic that ties them together, and the actions it takes. When someone messages the agent in Teams, n8n orchestrates the work across whatever systems are needed and returns the result right back into the conversation.
Microsoft 365 is where so much of daily work happens. With Agent 365, n8n users can build agents that don't just automate in the background, they participate directly in that work. And for Microsoft developers, n8n gives them a visual builder and over a thousand native integrations to connect agents to the rest of their tech stack.
— Ben Kiziltug, Vice President, n8n
This builds on the collaboration announced at Ignite in November 2025. With Agent 365 now generally available, the full capabilities are accessible to any organization with a Microsoft 365 tenant and n8n. n8n is available at the Microsoft Marketplace for teams that want to deploy within their Azure environment. That means governed access to Agent 365's MCP tools directly from the n8n canvas, and the ability to deploy agents that show up as members of the team inside Microsoft apps.
We're happy to welcome n8n to the Agent 365 ecosystem. With Agent 365, partners like n8n can bring agents directly into the flow of work across Microsoft 365. It gives developers a secure, governed way to build agents that show up where people already collaborate, communicate, and create.
— Srini Raghavan, Corporate Vice President, Microsoft Copilot & Agents Ecosystem at Microsoft Corp.
Learn how to build an agent in n8n using the Microsoft Agent 365 Trigger node, connect it to the tools your team relies on, and put it to work inside Microsoft Teams, Outlook, Word, and SharePoint. Live demo and Q&A.
What you can build
Consider a customer support lead who starts each morning checking Zendesk for overnight escalations, pulling account context from Salesforce, reviewing PagerDuty for any open infrastructure issues that might affect customers, and then manually posting a summary in Teams and following up in email with the account team. With an n8n-powered Agent 365 agent, she @mentions the agent in a Teams channel and it handles the rest. The agent pulls the tickets, cross-references account data, and checks for any active incidents across three systems before posting a chat message and drafting an email with next steps and owner assignments.
Coordinating across services is straightforward on the n8n canvas by adding them as tools for the Microsoft Agent 365 node. Zendesk is connected to the agent as a drag and drop node with its own credentials that can be configured once and reused across workflows. Salesforce, PagerDuty, each is added the same way. The agent calls whichever tools it needs based on the conversation and the context provided in its system prompt. The Microsoft services, like sending emails and posting in Teams, come through Agent 365's MCP servers and are enabled through a toggle in the Microsoft Agent 365 node configuration.
Existing n8n workflows and third party MCP tools make this even more powerful. Say that support lead built an n8n workflow that enriches a customer record by pulling in data from Salesforce, cross-checking contract status, and scoring the account's health. That workflow can be packaged as a sub-workflow and become a tool the agent calls. She doesn't need to rebuild that logic inside the agent. She just connects to it. If the team uses Datadog for observability, its MCP server can be connected as a tool through the MCP Client node, giving the agent real-time visibility into service performance when triaging customer issues.
Design visually, code when you need to, govern everything
When you need to go beyond the convenience of the built-in nodes, n8n lets you drop into JavaScript or Python at any point in the workflow. Use this option for custom data transformations and business logic, or call internal services through an HTTP request node. Write code where it makes sense, use visual nodes everywhere else, and combine them freely in the same workflow.
Governance works across two layers. Agent 365 handles identity, lifecycle management, security and compliance at the Microsoft level: Entra ID, conditional access policies, data protection through Purview, and threat detection through Defender, as well as agent registry, analytics, and policies in the Microsoft 365 Admin Center. n8n handles governance at the workflow level including RBAC for workflow access, credential management for connected services, guardrails for validating inputs and running safety checks like PII detection, error handling workflows if something goes wrong. Both layers work together so the agent is governed from identity all the way through to execution. If you have data residency or compliance requirements, n8n can be self-hosted on your own infrastructure giving you full control over where your data lives and how it's processed.
For teams that have struggled to get visibility into what’s happening in AI agents under the hood, building in n8n means you can see what the agent actually does at runtime. Every execution produces a detailed record of the agent's output and behavior, including each decision it made, every tool it called, and the data returned from those tools. The entire chain is visible, testable, and editable at any step.
Get started
If you already use n8n with Microsoft services through individual nodes, those workflows still work. Agent 365 is a new path for teams that want their agents to show up inside Microsoft apps and interact like a member of the team.
Getting an Agent 365 agent running involves some initial setup on the Microsoft side: registering the app in Entra ID, creating an agent blueprint through the Agent 365 CLI, and getting admin approval for the agent within your tenant. On the n8n side, add the Agent 365 Trigger node and connect your tools. Native nodes, MCP-connected services, and sub-workflows all work as tools the agent can call. Once the agent is approved, it shows up in Teams, Outlook, SharePoint, and Word, ready for your team to @mention and put to work.
The Microsoft Agent 365 Trigger node is available now in n8n. See n8n docs to set up and build your agent, and check Microsoft's Agent 365 documentation for how to provision for your organization.