API plugins for Microsoft 365 Copilot
Product
확장 모델로, Copilot이 실행 시점에 외부 API를 실시간으로 호출하여 보통 티켓 생성이나 레코드 업데이트 같은 작업을 수행합니다. 해당 데이터는 Microsoft 365에 인덱싱되지 않습니다.
Description
API plugins for Microsoft 365 Copilot are an extensibility model in which Copilot calls a live external API at runtime, typically to perform an action such as creating a ticket or updating a record. The data behind the API is not indexed into Microsoft 365. Instead, the API stays hosted outside Microsoft 365 and is queried or modified on demand when a user asks for it in natural language.
According to the Microsoft Learn overview, API plugins are custom actions that let declarative agents in Microsoft 365 Copilot work with REST APIs that have an OpenAPI description. They are not standalone installable packages. Only the agent that contains a plugin can use its functions. Users can ask Copilot to read data and, if the API permits it, to create, update or delete data.
The mechanism rests on two descriptive artifacts. The plugin manifest tells Copilot how to discover the API, and the OpenAPI description supplies operations and parameters. Copilot reads the descriptions of the OpenAPI operations and parameters to decide which operation fits a request and how to map the user's input to parameters. Because the selection depends on these descriptions, clear and accurate wording in the OpenAPI definition affects how reliably Copilot invokes the right endpoint. Adaptive Card templates in the manifest control how results are displayed, and citation metadata can be taken from the API response.
Authentication and consent are part of the model. If the API requires authentication, the plugin obtains a token or key from a token store, and the agent asks the user to sign in when needed. Copilot requests consent on first use. After that, read operations run without confirmation, while write operations still ask the user first. Developers can override these defaults.
The documentation also names practical limits that influence design decisions. Up to five plugins in an agent manifest are always added to the prompt; beyond that, a semantic selection based on plugin descriptions applies. Quality can drop from about ten functions, and the token window truncates large inputs and outputs. Plugin authors therefore have reason to keep the set of functions focused and the responses compact. The documentation names no licensing requirements for API plugins.
The M365.FM episode "Build Custom Microsoft 365 Copilot Plugins" (14 August 2025) covers the same idea from a developer angle. The episode says that by providing an OpenAPI definition, developers enable Copilot to invoke API endpoints reliably, map parameters and handle responses, which makes it easier to build Copilot extensions and API plugins for Microsoft 365 Copilot. It also states that Power Automate flows and orchestrator patterns can be invoked by Copilot extensions to automate multi-step processes, so that Copilot can trigger business workflows, update systems and return results conversationally.
In practice, API plugins fit situations where the source of truth lives in another system and the goal is an action or a current lookup rather than a searchable copy of the content. Typical examples are opening a ticket in a service desk system or updating a record in a line-of-business application. Teams weighing this model against approaches that index data into Microsoft 365 should consider that API plugins always depend on the availability and behavior of the external API at the moment of the request.
Overview compiled with AI from M365.FM episodes and the facts on this page. You can improve it with the edit button below.
| Type | Product |
|---|---|
| Documentation | API plugins for Microsoft 365 Copilot – Microsoft Learn learn.microsoft.com |
Articles about API plugins for Microsoft 365 Copilot (2)
API plugins for Microsoft 365 Copilot을(를) 다룬 글:
- Copilot 플러그인, Graph 기반 응답, Copilot Studio 커넥터 설명
· Mirko Peters
Microsoft 365 Copilot에서 플러그인, API 플러그인, Microsoft Graph 기반 응답(grounding), Copilot 커넥터, Copilot Studio 커넥터가 각각 무엇을 의미하는지, 서로 어떻게 다른지, 그리고 어떤 작업에 어떤 것을 써야 하는지 설명합니다. - Copilot Plugins, Graph Grounding and Copilot Studio Connectors Explained
· Mirko Peters
What plugins, API plugins, grounding through Microsoft Graph, Copilot connectors and Copilot Studio connectors each mean in Microsoft 365 Copilot, how they differ, and which one to
Frequently asked questions
Questions marked with AI were written with AI from the facts on this page.
API 플러그인과 Microsoft 365 Copilot의 인덱싱된 데이터는 어떤 차이가 있습니까? (AI)
API 플러그인을 사용하면 Copilot은 실행 시점에 외부 API를 실시간으로 호출하며, 데이터는 Microsoft 365에 인덱싱되지 않습니다. API는 Microsoft 365 외부에서 호스팅되며 필요에 따라 조회되거나 변경됩니다.
API 플러그인이 작동하려면 무엇이 필요합니까? (AI)
OpenAPI 설명이 포함된 REST API와 플러그인 매니페스트가 필요합니다. Copilot은 작업 및 매개 변수 설명을 사용하여 호출할 작업을 선택하며, 플러그인은 선언형 에이전트의 일부여야 합니다.
Copilot은 API 플러그인을 사용하기 전에 확인을 요청합니까? (AI)
Copilot은 처음 사용할 때 동의를 요청합니다. 이후에는 읽기 작업이 확인 없이 실행되는 반면 쓰기 작업은 여전히 확인을 요청하며, 개발자는 이러한 기본값을 재정의할 수 있습니다.
Community wiki
Short info above, community knowledge below: add practical tips, examples or corrections. New contributions are reviewed first; trusted members publish directly.
Experts and companies for this topic
API plugins for Microsoft 365 Copilot에 전문성이 있는 커뮤니티 멤버:
- Mirko Peters · Microsoft MVP
Sources
- API plugins for Microsoft 365 Copilot – Microsoft Learn learn.microsoft.com
- RSS 피드: API plugins for Microsoft 365 Copilot
- Articles on this topic: 2
Maintained by Mirko Peters (Microsoft MVP) · Last updated: