Explore Microsoft 365 Agents SDK
Overview of Microsoft 365 Agents SDK and Its Components
Microsoft 365 Agents SDK refers to the evolving ecosystem of tools and frameworks for building AI-powered “Copilot” agents across the Microsoft 365 platform. Its key components include the Microsoft 365 Agents SDK, the Teams AI Library, and the Microsoft 365 Agents Toolkit, each serving a distinct role in the development process:
Microsoft 365 Agents SDK (Pro-Code Framework): The Agents SDK is a code-first framework for developers to create enterprise-grade, scalable, multi-channel agents. It’s often described as the evolution of the Bot Framework for the generative AI era. With the Agents SDK, you can build agents in languages like C#, JavaScript/TypeScript, or Python that leverage any LLM or AI service of your choice (e.g. Azure OpenAI, Azure AI Foundry, etc.) and deploy them to various channels – including Microsoft Teams, Outlook (email), web chat, SMS (via Twilio), and even third-party platforms like Slack or Discord. Unlike older bot development models, the Agents SDK provides full control over orchestration and integration, meaning you can plug in custom and incorporate multiple data sources. It uses Azure Bot Service and follows an activity-based model (similar to Bot Framework), managing turns (conversational exchanges) and activities (events like messages) so you can focus on implementing your agent’s intelligence.
Teams AI Library (Teams-Focused AI SDK): The Teams AI Library is a specialized SDK tailored for building AI-driven bots and “custom engine” agents within Microsoft Teams. It provides higher-level, Teams-specific abstractions that simplify conversational bot development by reducing boilerplate code and integrating natively with Teams features. For example, Teams AI Library includes a built-in Action Planner (an orchestrator that plans the bot’s actions in response to user input) and supports GPT-based language models (via Azure OpenAI or OpenAI API) out-of-the-box. This library brings advanced capabilities such as multi-agent communication (Agent-to-Agent, A2A) – allowing agents to coordinate with each other – and the Model Context Protocol (MCP) for sharing context among AI components. Essentially, Teams AI Library is ideal when building collaborative agents that operate in Team chats, channels, or meetings, where they might need to maintain conversation context, work alongside users in real-time, and present rich content (it has native support for Adaptive Cards in Teams). It’s a modernized successor to the classic Bot Framework for Teams bots, consolidating previous tools (Bot Builder SDK, Graph APIs, etc.) into one cohesive library.
Microsoft 365 Agents Toolkit (Development Tools): The Agents Toolkit is a suite of developer tools and services that streamlines building and deploying these agents. It is essentially an evolution of the “Teams Toolkit” and expands it for Copilot and multi-channel agent development. The Toolkit includes:
- Visual Studio Code Extension & Visual Studio Extension: which provide project templates (for both custom engine agents and declarative agents), graphical interface to configure your agent, one-click debugging, and integrated commands for provisioning and deployment.
- CLI (Command Line Interface): a command-line tool for automating tasks like manifest generation, resource provisioning on Azure, and CI/CD integration. It helps manage multiple environments (dev/test/prod) and can sideload and register your bot to Teams or other channels with ease
- Agents Playground: an integrated local testing environment that simulates the Microsoft Teams client and Copilot chat, allowing you to run your agent locally without needing to deploy to a cloud or have a tenant set up. The Playground supports simulating user messages, testing adaptive cards, and even mock data or triggers, which greatly speeds up iterative development.
- The toolkit provides a “write once, run everywhere” approach: you can build your agent once and use the Toolkit to deploy it to multiple front-ends (Teams, Outlook, Office add-ins, the Copilot in Microsoft 365 app, etc.) from a central project. It also simplifies implementing cross-cutting concerns like SSO authentication, storage integration, and creating CI/CD pipelines (with pre-made GitHub Actions/Azure DevOps tasks) for your agent.
In summary, the Microsoft 365 SDK components work together to enable two main development paradigms for Copilot-like agents: a pro-code approach (using the Agents SDK + Teams AI Library with the Agents Toolkit for support) for maximum flexibility and multi-channel reach. An enterprise developer can use the Agents SDK to build a highly customized agent (with custom orchestration and AI logic) and rely on the Agents Toolkit to scaffold the project, test it in the Playground, and deploy it to Azure and Microsoft 365 channels. If the target is specifically a conversational experience inside Teams, the Teams AI Library can be used within that project to leverage built-in AI conversation features.