Skip to content
Dhanasvi
Sign inExplore
Agent Framework logo

Agent Framework

Verified

Open framework for building production AI agents and workflows in Python and .NET.

Autonomous AgentsAgent Frameworks 11.3kOpen source
View on GitHub
Updated 2026-06-15
Agent Framework GitHub repository

What is Agent Framework?

Microsoft Agent Framework is an open-source toolkit for developing durable AI agents and multi-agent applications. It offers native support for both Python and C#/.NET with matching APIs, allowing developers to define agents, connect them to various models, and compose them into workflows using patterns like sequential, concurrent, or handoff execution.

Core capabilities include checkpointing for restartability, OpenTelemetry tracing, YAML-based declarative definitions, and optional hosting on Foundry infrastructure. Skills let agents draw from files or code libraries, while middleware handles custom processing and human-in-the-loop steps.

It suits engineering teams that need governance, monitoring, and provider flexibility without locking into a single runtime or cloud service.

Capabilities

build ai agents
orchestrate agents
run ai agents
use workflow primitives

What you can build with Agent Framework

Production multi-agent orchestration

Design graph workflows that coordinate several agents with built-in checkpointing and streaming for reliable execution at scale.

Enterprise observability and control

Add distributed tracing and human oversight to agent runs while maintaining governance across development and deployment stages.

Cross-language agent deployment

Build once and run agents in both Python and .NET environments with consistent behavior and easy migration between providers.

Install Agent Framework

Install
pip install agent-framework
Quick start
pip install agent-framework
# This will install all sub-packages, see `python/packages` for individual packages.
# It may take a minute on first install on Windows.
  1. 1Choose Python or .NET based on your stack and install the corresponding package.
  2. 2Create a basic agent instance and connect it to an LLM provider such as Azure OpenAI or OpenAI.
  3. 3Define a simple workflow using sequential or concurrent patterns and add any required middleware.
  4. 4Test the agent locally with the DevUI or run sample scripts to verify behavior and observability.
  5. 5Deploy to Foundry or your chosen environment and enable checkpointing for production use.

Agent Framework: pros & cons

Pros

  • +Strong production features including durability, observability, and human-in-the-loop support
  • +Consistent APIs across Python and .NET reduce context switching for mixed teams
  • +Flexible orchestration patterns and middleware allow complex workflows without heavy custom code
  • +Broad provider support keeps architecture choices open as needs change

Cons

  • –Requires familiarity with workflow concepts that may feel heavy for simple single-prompt use cases
  • –Microsoft-centric hosting options could influence long-term platform decisions
  • –Experimental labs features may change frequently during early adoption
Did you find this helpful?

Frequently asked questions

No, it works with OpenAI, Azure OpenAI, and other providers while offering optional Foundry hosting.

User reviews

Verified reviews from the community shape this listing's rating.

Loading reviews…

Sign in to review

Promote Agent Framework

Add this badge to your website, or share the tool.

DFeatured on DhanasviAgent Framework 1