agent-framework
Microsoft Agent Framework (MAF) is an open, multi-language framework for building production-grade AI agents and multi-agent workflows in Python and .NET, with a Go SDK in a separate repository. Microsoft positions it as the successor to AutoGen, which is now in maintenance mode, and it is built for teams taking agents from prototype to production.
It is a strong fit when you need orchestration beyond a single prompt: graph-based workflows with sequential, concurrent, handoff and group-collaboration patterns, checkpointing, streaming and human-in-the-loop control. It also offers middleware for request and response processing, built-in OpenTelemetry tracing, declarative agents defined in YAML, agent skills, an interactive DevUI for debugging, experimental "Labs" packages for benchmarking and reinforcement learning, and the option to host agents on Microsoft Foundry with a couple of lines of code. Multiple model providers are supported.
Getting started: Python users run `pip install agent-framework`; .NET users add the `Microsoft.Agents.AI` package (plus Foundry packages for the hosted quickstart). Teams on AutoGen should follow Microsoft's official migration guide.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-24
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 22
- Latest release
- dotnet-1.26.0 (2026-10-23)
- Open issues
- 849
- Recent issues with at least one comment
- 10/17
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #4842 .NET: Discussion: Agent Identity and Trust for Multi-Agent Workflows17 comments · 2026-03-22
- #2084 .NET: DevUI limitation should be documented and addressed if possible16 comments · 2025-11-11
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- You are starting a new Microsoft-ecosystem agent project in Python or .NET and want the official successor to AutoGen.
- You need built-in workflow checkpointing, middleware, OpenTelemetry tracing, declarative YAML agents, and a DevUI under MIT.
- You only need a small, single-file Python agent without a multi-package framework; Pydantic AI is lighter.
- You rely on Chinese LLM providers and realtime voice pipelines out of the box; AgentScope has dedicated support for those.
How it differs
vs Microsoft AutoGen★ 61.6kMicrosoft Agent Framework is the active, production-focused successor to AutoGen. AutoGen is now in maintenance mode and no longer receives new features.
vs adk-python★ 22.0kMicrosoft Agent Framework offers first-class .NET and Python packages with Azure/Foundry integration; Google ADK is Apache-2.0 and targets Google Cloud and Gemini workflows.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!