OSS Tanbou
Sign in with GitHub
View on GitHubProject website

Build multi-stage AI workflows by coordinating specialized agents through conversations and events

OSS scale score 333.3OSS health 62Update pace 0 commits/30d
About these scores

OSS scale score is an unbounded metric that log-compresses and weights Stars, Watchers, Forks, and Contributors. Discovery score is the current OSS scale score minus the score at discovery. Update pace is commits in the last 30 days, growth momentum is the OSS scale score difference within the recent observation window, and OSS health is a 0–100 rating based on available recency, Community Health, and release data.

Stars
61,254
Primary language
Python
License
CC-BY-4.0
Repository last updated
Apr 15, 2026
On this page

Overview

AutoGen is a framework for applications in which multiple AI agents act autonomously or collaborate with people. Projects can choose among the low-level Core API, the more opinionated conversational AgentChat API, and the Extensions API for models, code execution, and other integrations. Python and .NET implementations, the AutoGen Studio interface, and AutoGen Bench evaluation tooling sit in the same ecosystem.

Features and best fit

Based on official documentation; not hands-on tested · Content checked:

Key features

Choose an abstraction level from runtimes to conversational APIs

The Core API provides message passing, event-driven agents, and local or distributed runtimes. AgentChat builds on it with higher-level patterns such as two-agent conversations and group chats, allowing teams to choose between faster composition and detailed execution control.

Sources: [2]

Connect models, code execution, and MCP through extensions

The Extensions API supplies implementations for model clients such as OpenAI and Azure OpenAI, code execution, and MCP connectivity. Teams can also explore configurations visually in AutoGen Studio and evaluate behavior across tasks with AutoGen Bench.

Sources: [2][3]

Best fit

Fits AI workflows with explicit roles and termination rules

It is useful for prototyping research, planning, execution, and review as separate agents, or for applications that insert human approval into a multi-stage process. Teams can validate the conversation flow with a higher-level API and move only the necessary parts down to Core.

Sources: [2]

Before adoption

Control model cost, tool authority, and termination in the application

Multi-agent loops increase model calls and latency, so applications need explicit iteration limits, budgets, timeouts, and evaluation criteria. Code execution and MCP servers require isolation and least-privilege access, and untrusted model output must not flow directly into actions. The Python packages require 3.10 or newer. GitHub detects CC-BY-4.0 at the repository root, while AgentChat code includes an MIT LICENSE-CODE; verify the terms for the specific files and artifacts being used or redistributed.

Sources: [2][4][5][6]

Official sources

  1. [1]microsoft/autogen repository metadata(2026-10-04)
  2. [2]AutoGen python-v0.7.5 README(2026-10-04)
  3. [3]AutoGen python-v0.7.5 release(2026-10-04)
  4. [4]AutoGen AgentChat 0.7.5 package metadata(2026-10-04)
  5. [5]AutoGen repository CC-BY-4.0 license(2026-10-04)
  6. [6]AutoGen AgentChat MIT code license(2026-10-04)
Supplemental curator note

Adding agents is not an objective by itself. Map the boundaries a single process cannot handle, failure termination rules, and each tool’s authority first, then measure both call volume and output quality.

Try it in 3 steps

  1. 1

    Create an isolated virtual environment

    Use Python 3.10 or newer, isolate AutoGen dependencies, and continue in the same terminal and working directory.

    python -m venv .venv && source .venv/bin/activate
  2. 2

    Install the reviewed AgentChat release

    Install only the pinned AgentChat 0.7.5 package. This step needs no model-provider SDK, API key, or paid request.

    python -m pip install "autogen-agentchat==0.7.5"
  3. 3

    Verify the AgentChat API import

    Import the main agent class without connecting to a model. A real conversation needs the relevant extension package, provider API key, and network access and may incur provider charges. This step does not validate agent execution or response quality.

    python -c "from autogen_agentchat.agents import AssistantAgent; print(AssistantAgent.__name__)"
Check the official README

Growth

Growth trends · Last 30 days

61,254 Stars

Trend data is still being collected.

Development activity

Last 90 days · weekly

Commits (last 30 days)
0
Open PRs
539

Development activity is still being collected.

Built with

Categories and tags

Categories

Tags

GitHub data

GitHub dataView detailed GitHub data
View on GitHub

GitHub Topics

  • chatgpt
  • llm-agent
  • llm-framework
  • agentic
  • agentic-agi
  • agents
  • ai
  • autogen
  • framework
  • autogen-ecosystem
Stars
61,254
Forks
9,281
Watchers
530
Open issues
560
Contributors
534
Owner type
Organization
Primary language
Python
License
CC-BY-4.0
Repository last updated
Apr 15, 2026
Write a related article

Share a guide or use case for this OSS in Markdown. Articles are published after administrator approval.

Report incorrect information

Tell us if any listing information is incorrect or outdated.

After reading this page, do you know what to do next?