Python framework for building AI agents that use LLMs and tools, and coordinate with each other in multi-agent workflows.
microsoft/agent-framework
An open framework for building, orchestrating and deploying production-grade AI agents and multi-agent workflows in .NET and Python.
- Stars
- 14,565
- Forks
- 2,589
- Open issues
- 794
- Last commit
- 20 hours ago
What is it?
Microsoft Agent Framework is a multi-language SDK for building AI agents and multi-agent workflows in Python and C#/.NET, with a consistent API across both. It targets teams moving agents from prototype to production, covering orchestration beyond a single prompt or stateless chat loop.
Workflows are graph-based and support sequential, concurrent, handoff, and group collaboration patterns, along with checkpointing, streaming, human-in-the-loop control, and time-travel. The framework includes middleware for request and response processing, OpenTelemetry integration for tracing and monitoring, and declarative agent definitions in YAML. It works with providers such as Microsoft Foundry, Azure OpenAI, OpenAI, and the GitHub Copilot SDK.
The repository also includes samples for local development and cloud hosting, including Foundry hosted agents, A2A, and a DevUI for testing and debugging workflows. A separate Go SDK lives in its own repository.
Key features
- 01Consistent APIs across Python and C#/.NET
- 02Graph-based workflows with sequential, concurrent, handoff, and group patterns
- 03Checkpointing, streaming, and human-in-the-loop control
- 04Middleware pipeline for request, response, and exception handling
- 05Built-in OpenTelemetry tracing and monitoring
- 06Declarative agent definitions in YAML
Use cases
- Building multi-agent systems that hand tasks between specialized agents
- Running agent workflows that need durability, restartability, and human approval steps
- Deploying agents to Microsoft Foundry hosted infrastructure
- Adding observability to LLM-based agents in existing .NET or Python services
Quick start
- 1
Install the Python package
pip install agent-framework - 2
Authenticate with Azure CLI before running the sample
az login - 3
Save the Python quickstart code as a file and run it with Foundry endpoint and model configured
Things to know
- The quickstart requires a Microsoft Foundry project and Azure sign-in.
- Some providers and features depend on Azure or third-party services that you pay for and configure yourself.
- Third-party systems are used at your own risk under separate license terms.
In collections
Topics
FAQ
- What is agent-framework?
- An open framework for building, orchestrating and deploying production-grade AI agents and multi-agent workflows in .NET and Python. Microsoft Agent Framework is a multi-language SDK for building AI agents and multi-agent workflows in Python and C#/.NET, with a consistent API across both. It targets teams moving agents from prototype to production, covering orchestration beyond a single prompt or stateless chat loop.
- Is agent-framework free to use? What license does it have?
- agent-framework is open source under the MIT license. Check the license terms before using it in commercial products.
- Is agent-framework still maintained?
- Actively maintained. The last commit to agent-framework was on Nov 6, 2026.
- What can I use agent-framework for?
- Building multi-agent systems that hand tasks between specialized agents Running agent workflows that need durability, restartability, and human approval steps Deploying agents to Microsoft Foundry hosted infrastructure Adding observability to LLM-based agents in existing .NET or Python services
- How do I install agent-framework?
- Install the Python package: pip install agent-framework Authenticate with Azure CLI before running the sample: az login Save the Python quickstart code as a file and run it with Foundry endpoint and model configured
- Where can I get agent-framework?
- The source code and installation instructions are on GitHub: https://github.com/microsoft/agent-framework
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source Python framework for building production-ready LLM applications with modular pipelines and agent workflows for RAG, search and agents.
Python framework for creating simulations populated with AI agents that talk, act, remember and take custom actions in shared environments.
A multi-agent plugin for OpenCode that routes planning, codebase scouting, UI work and implementation to specialized agents, each on a model you choose.
Code-first Python framework from Google for building, evaluating, and deploying AI agents, optimized for Gemini but model-agnostic.
Self-hosted TypeScript agent runtime for coordinating multiple LLM agents, with human approval gates and offline-inspectable run journals.