Installation

Complete guide to installing AutoGen for Python and .NET

Installation

AutoGen is available for both Python and .NET. This guide covers installation for both platforms, prerequisites, and verification steps.

Python Installation

AutoGen requires Python 3.10 or later.

Prerequisites

1. Check Python version

   Verify you have Python 3.10 or later:

   python --version
   # or
   python3 --version
   

   If you need to upgrade, download from python.org.

2. Create a virtual environment (recommended)

   Using a virtual environment keeps AutoGen dependencies isolated:

   venv

   Linux/Mac

   python3 -m venv .venv
   source .venv/bin/activate
   

   python -m venv .venv
   .venv\Scripts\activate.bat
   

   python -m venv .venv
   .venv\Scripts\Activate.ps1
   

   To deactivate later:

   deactivate
   

   conda

   If you prefer conda, first install Conda.

   conda create -n autogen python=3.12
   conda activate autogen
   

   To deactivate later:

   conda deactivate
   

Core Packages

AutoGen has three main packages that can be installed separately or together:

autogen-core

Event-driven foundation with agent runtime, message passing, and distributed execution.

autogen-agentchat

High-level API with preset agents and team patterns. Built on autogen-core.

autogen-ext

Extensions for model clients, tools, and integrations.

Installation Options

Recommended: AgentChat + Extensions

For most users, install AgentChat with model client extensions:

pip install -U "autogen-agentchat" "autogen-ext[openai]"

This installs:

• autogen-core (dependency)
• autogen-agentchat for high-level agents and teams
• autogen-ext[openai] for OpenAI/Azure OpenAI support

Core API Only

For advanced users who want event-driven control:

pip install -U "autogen-core"

You’ll likely also want extensions:

pip install -U "autogen-ext[openai]"

Studio (No-Code GUI)

Install AutoGen Studio for visual workflow building:

pip install -U "autogenstudio"

Launch the UI:

autogenstudio ui --port 8080

Then open http://localhost:8080 in your browser.

Extension Options

The autogen-ext package supports many optional dependencies. Install only what you need:

Model Clients

# OpenAI and Azure OpenAI
pip install "autogen-ext[openai]"

# Azure with AAD authentication
pip install "autogen-ext[azure]"

# Anthropic Claude
pip install "autogen-ext[anthropic]"

# Google Gemini
pip install "autogen-ext[gemini]"

# Local models via Ollama
pip install "autogen-ext[ollama]"

# llama.cpp support
pip install "autogen-ext[llama-cpp]"

Tools & Capabilities

# Code execution in Docker
pip install "autogen-ext[docker]"

# Jupyter kernel code execution
pip install "autogen-ext[jupyter-executor]"

# MCP server integration
pip install "autogen-ext[mcp]"

# Web browsing capabilities
pip install "autogen-ext[web-surfer]"

# File handling tools
pip install "autogen-ext[file-surfer]"

Memory & Storage

# ChromaDB vector store
pip install "autogen-ext[chromadb]"

# Redis caching
pip install "autogen-ext[redis]"

# DiskCache
pip install "autogen-ext[diskcache]"

# Mem0 memory system
pip install "autogen-ext[mem0]"

Advanced

# gRPC for distributed runtime
pip install "autogen-ext[grpc]"

# GraphRAG integration
pip install "autogen-ext[graphrag]"

# LangChain integration
pip install "autogen-ext[langchain]"

# Semantic Kernel
pip install "autogen-ext[semantic-kernel-core]"

# Rich terminal output
pip install "autogen-ext[rich]"

You can combine multiple extensions: pip install "autogen-ext[openai,docker,chromadb]"

Verify Python Installation

Test that packages are installed correctly:

import autogen_core
import autogen_agentchat
import autogen_ext.models.openai

print(f"autogen-core: {autogen_core.__version__}")
print(f"autogen-agentchat: {autogen_agentchat.__version__}")
print(f"autogen-ext: {autogen_ext.__version__}")
print("\nAutoGen installed successfully!")

Latest stable version: 0.7.5 (check releases for updates)

.NET Installation

AutoGen for .NET supports both the legacy AutoGen.* packages and the new Microsoft.AutoGen.* packages.

Prerequisites

1. Install .NET SDK

   Download and install .NET 8.0 SDK or later.

   Verify installation:

   dotnet --version
   
2. Create a new project

   dotnet new console -n MyAutoGenApp
   cd MyAutoGenApp
   

Package Options

AutoGen for .NET has two package families:

• Microsoft.AutoGen.* (new, event-driven model) - Recommended
• AutoGen.* (legacy, v0.2 style) - Being gradually deprecated

This guide focuses on the new packages.

Microsoft.AutoGen (Recommended)

The new event-driven packages:

# Core contracts and types
dotnet add package Microsoft.AutoGen.Contracts

# Agent runtime
dotnet add package Microsoft.AutoGen.Core

# gRPC support for distributed runtime
dotnet add package Microsoft.AutoGen.Core.Grpc

# Runtime gateway
dotnet add package Microsoft.AutoGen.RuntimeGateway.Grpc

See .NET Core documentation for usage.

AutoGen (Legacy)

The older packages for AutoGen v0.2 style:

# Core abstractions
dotnet add package AutoGen.Core

# OpenAI integration
dotnet add package AutoGen.OpenAI

# Source generators for type-safe functions
dotnet add package AutoGen.SourceGenerator

These packages will continue to receive bug fixes and security patches but new features go to Microsoft.AutoGen.*

Sample .NET Code

Here’s a simple agent using the legacy packages:

using AutoGen;
using AutoGen.OpenAI;

var openAIKey = Environment.GetEnvironmentVariable("OPENAI_API_KEY") 
    ?? throw new Exception("Please set OPENAI_API_KEY");

var gpt35Config = new OpenAIConfig(openAIKey, "gpt-3.5-turbo");

var assistantAgent = new AssistantAgent(
    name: "assistant",
    systemMessage: "You are a helpful AI assistant.",
    llmConfig: new ConversableAgentConfig
    {
        Temperature = 0,
        ConfigList = new[] { gpt35Config },
    })
    .RegisterPrintMessage();

var userProxyAgent = new UserProxyAgent(
    name: "user",
    humanInputMode: HumanInputMode.ALWAYS)
    .RegisterPrintMessage();

await userProxyAgent.InitiateChatAsync(
    receiver: assistantAgent,
    message: "Hello, assistant!",
    maxRound: 10);

Verify .NET Installation

Run your project:

dotnet run

For more examples, see the .NET samples.

Development Installation

If you want to contribute to AutoGen or use the latest unreleased features:

From Source (Python)

git clone https://github.com/microsoft/autogen.git
cd autogen/python

# Install in development mode
pip install -e packages/autogen-core
pip install -e packages/autogen-agentchat
pip install -e "packages/autogen-ext[openai]"

Nightly Builds (.NET)

Add the nightly feed to your NuGet sources:

dotnet nuget add source https://pkgs.dev.azure.com/AGPublish/AGPublic/_packaging/AutoGen-Nightly/nuget/v3/index.json \
  --name AutoGen-Nightly

Then install packages from this feed:

dotnet add package Microsoft.AutoGen.Core --prerelease

Platform-Specific Notes

Windows

PowerShell Execution Policy

If you can’t activate virtual environments, run:

Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

Long Path Support

Enable long paths to avoid installation issues:

# Run as Administrator
New-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\FileSystem" -Name "LongPathsEnabled" -Value 1 -PropertyType DWORD -Force

macOS

Python Installation

Use Homebrew for the latest Python:

brew install python@3.12

M1/M2/M3 Macs

Some extensions may require Rosetta for x86 dependencies:

softwareupdate --install-rosetta

Linux

System Dependencies

Some extensions need system packages:

Ubuntu/Debian:

sudo apt-get update
sudo apt-get install -y python3-dev build-essential

CentOS/RHEL:

sudo yum install python3-devel gcc

Troubleshooting

pip install fails with 'No matching distribution'

This usually means your Python version is too old. AutoGen requires Python 3.10+:

python --version

Upgrade Python if needed, or use python3.10, python3.11, or python3.12 explicitly:

python3.11 -m pip install autogen-agentchat

Permission denied errors

On Linux/Mac, you might need --user flag:

pip install --user autogen-agentchat

Or use a virtual environment (recommended).

SSL certificate errors

Update certificates:

pip install --upgrade certifi

Or temporarily bypass (not recommended for production):

pip install --trusted-host pypi.org --trusted-host files.pythonhosted.org autogen-agentchat

Import errors after installation

Make sure you’re using the correct Python environment:

which python  # Linux/Mac
where python  # Windows

If using a virtual environment, ensure it’s activated.

Version conflicts

AutoGen has specific version requirements. If you get conflicts:

# Create a fresh environment
python -m venv fresh_venv
source fresh_venv/bin/activate  # or activate.bat on Windows
pip install autogen-agentchat

Upgrading AutoGen

Python

Update to the latest version:

pip install --upgrade autogen-agentchat autogen-ext

Check installed version:

pip show autogen-agentchat

.NET

Update packages:

dotnet add package Microsoft.AutoGen.Core

Or update all packages:

dotnet restore --force

Migrating from AutoGen v0.2? The new version has breaking changes. See the Migration Guide for detailed upgrade instructions.

Next Steps

Now that AutoGen is installed:

Quickstart

Build your first agent in 5 minutes

Core Concepts

Learn about agents, teams, and tools

Model Configuration

Set up OpenAI, Azure, or other providers

Examples

Explore sample applications

Additional Resources

• Python Package Index: pypi.org/project/autogen-agentchat
• .NET NuGet Gallery: nuget.org/packages/Microsoft.AutoGen.Core
• GitHub Repository: github.com/microsoft/autogen
• Release Notes: GitHub Releases

Having installation issues? Ask for help on GitHub Discussions or Discord.