An official open-source bilingual AI agent orchestration framework by Microsoft, supporting Python and .NET with graph workflows, multi-agent collaboration, and OpenTelemetry observability.
The Microsoft Agent Framework is an official open-source framework designed for building, orchestrating, and deploying AI agents and multi-agent workflows. Its core strength lies in supporting both Python and .NET (C#) with consistent API styles.
Core Features
- Graph-based Workflows: Supports advanced state management features like streaming, persistent checkpointing, time travel, and human-in-the-loop.
- Multi-Agent Collaboration: Includes built-in orchestration patterns such as Sequential, Concurrent, Group Chat, Handoff, and Magentic.
- Observability: Integrates OpenTelemetry for distributed tracing and monitoring, meeting production-grade requirements.
- Developer Tools: Provides a DevUI for interactive development, testing, and debugging of workflows.
- Multi-Model Support: Supports various LLM providers (e.g., OpenAI, Azure OpenAI) via adapters.
Architecture & Installation
The repository contains two main modules: python/ and dotnet/. It abstracts LLM calls via ResponsesClient and supports a flexible middleware pipeline.
- Python Install:
pip install agent-framework --pre(Requires Python ≥ 3.10) - .NET Install:
dotnet add package Microsoft.Agents.AI --prerelease
The project is open-sourced under the MIT license and is currently in active development (RC phase).