AutoGen
@pyautogen
Followers
10K
Following
1K
Media
179
Statuses
874
Free. Cross-language. Open-source. An ecosystem for AI agent research and applications from @Microsoft
Joined October 2023
🚀 Introducing Magentic-UI — an experimental human-centered web agent from @MSFTResearch . It automates your web tasks while keeping you in control 🧠🤝—through co-planning, co-tasking, action guards, and plan learning. 🔓 Fully open-source. We can't wait for you to try it. 🔗
6
68
264
🌻Deepseek R1 + @ollama + AutoGen. AutoGen provides essentials for building agents powered with @deepseek_ai thinking models. You can.- run agents locally and privately.- track/extract model's thought and content.- stream both thoughts and content to UI/application
8
81
474
📢Introducing Magentic-One, a generalist 5-agent multi-agent system for solving open-ended web- and file-based tasks. 🤖🤖🤖🤖🤖. Magentic-One represents a significant step towards agents that can complete tasks that people encounter in their daily lives and can achieve strong
3
112
424
🌻 DeepResearch + AutoGen. DeepResearch is agentic system from @OpenAI that uses LLMs to perform complex research. You can implement such systems using AutoGen. A DeepResearch example is now available in AutoGen Studio's gallery. This team performs deep research using .- web
9
51
324
🚨We just released a new #AutoGen tutorial . And with that getting started became even easier! First 5 chapter are already online to help you learn about. - agents that can converse.- termination.- adding humans in-the-loop.- code executors.- multi-agent patterns. Lets us know if
2
38
178
We are excited to announce a preview of the new architecture of AutoGen (coming in v0.4). To learn more, see.Blog: Pull request: Come help us shape the future of AutoGen!
5
57
181
🌻 Hugging Face Open DeepResearch + AutoGen. Hugging Face recently implemented an open source version of DeepResearch and used AutoGen's MagenticOne's agents and tools to build it. Kudos back to the @huggingface team! Open-source ftw!. "These tools were taken from the excellent
4
33
172
🌻 AutoGen Studio v0.4.1 is out. Implements many new exciting features.- specify agents/teams using JSON.- export existing agents/teams to JSON.- test teams in team builder .- new gallery with examples (DeepResearch & Web agents)
4
25
174
Build @pyautogen Agents without writing *any* code. AutoGen studio is a low code / no code tool for prototyping agents in AutoGen. You can do the following:. - Use the team builder to compose teams (drag and drop from a built in gallery of teams including a deep research agent
2
22
162
📰: #AutoGen recently won the best paper award at the LLM Agent Workshop at ICLR-2024! The workshop had at 90+ accepted oral presentations and posters.
3
26
149
This new paper from @MIT uses multiple #AutoGen agents to generate, refine, and validate research hypothesis!. SciAgents: Automating Scientific Discovery Through Multi-Agent Intelligent Graph Reasoning.by Ghafarollahi & @ProfBuehlerMIT . "We present SciAgents, an approach that
5
38
130
🌻GraphRAG + AutoGen. GraphRAG takes RAG to the next level -- instead of plain text snippets, it builds knowledge graphs for smarter retrieval and generation. You can now use GraphRAG as a tool for AutoGen agents.
0
29
110
HUGE: The biggest upgrade to #AutoGen just dropped!. v0.4 (stable) is finally here. For detail, checkout the blog below.
Announcing AutoGen 0.4, fully reimagined library for building advanced agentic AI systems, developed to improve code quality and robustness. Its asynchronous, event-driven architecture is designed to support dynamic, scalable workflows. Learn more:
5
34
109
🌻 Use MarkitDown with Claude Desktop. We recently added a basic STDIO and SSE MCP server that allows users to use MarkitDown's ability to convert files into markdown that LLMs can consume. See microsoft/markitdown on GitHub for details
3
26
109
you know whats even wilder?! 🤣. you can pair agents in even more sophisticated ways . - autogen is open source with MIT license.- provides pre-built high performing agents.- actor model for agent-agent communication.- use python, dotnet, or low code studio. (here's a team
this is wild 🤯🤯🤯. just paired openai operator with replit agent to build an app. watch as two ai agents team up, exchange credentials, and start testing . ai agent 🤝 ai agent. the future is here and it’s insane!
4
14
98
🚀 AutoGen 0.4 is a from-the-ground-up rewrite, bringing new features and improvements. Learn more about its journey and future plans in our latest blog post! We're gathering community feedback before the official release. 📣 For prior users, AutoGen 0.2 is still actively.
We are excited to announce a preview of the new architecture of AutoGen (coming in v0.4). To learn more, see.Blog: Pull request: Come help us shape the future of AutoGen!
2
14
88
🧵#AutoGen Magentic-One is exhibits strong generalist performance, you can use it for so many open-ended, every day tasks. Here is a thread of videos for a wide variety of every day tasks! To learn more about Magentic-One, checkout the original release tweet below. task: find
📢Introducing Magentic-One, a generalist 5-agent multi-agent system for solving open-ended web- and file-based tasks. 🤖🤖🤖🤖🤖. Magentic-One represents a significant step towards agents that can complete tasks that people encounter in their daily lives and can achieve strong
2
15
86
New open-source Python package: markitdown . This package makes it easy to convert various files to Markdown (e.g., for indexing, text analysis, etc.). #AutoGen uses it to enable file handling in SOTA agentic systems like #MagenticOne. The API is super simple. cc @adamfourney
2
19
84
🌻 AI Agents for Beginners. Checkout this completely free new course on agents by @Microsoft .- use LLMs @github models.- learn about AutoGen and Semantic Kernel.
The "AI Agents For Beginners" course is now live! . - 10 Lessons available today teaching you the basics of building AI Agents. - Code Samples using @github Models (free)w/ @pyautogen and Semantic Kernel.- Translations in 9 Different Languages.
3
17
77
🌻Web agents + AutoGen. Many websites don’t provide APIs, which makes it hard to automate tasks like filling out forms, gathering information, or interacting with content. Tools that can operate a browser directly solve this by mimicking human actions like clicking and typing.
5
17
74
Creating multi-agent applications just got even easier with AutoGen and Ollama!. Tyvm @ollama 🤝.
Exciting update from Ollama! 🚀This update on OpenAI compatibility also makes it extremely easy to use local models in AutoGen, as featured in this blog post! @pyautogen + @ollama 🚀🚀
1
17
66
📹 AutoGen Studio + @ollama + @GroqInc + @kaggle . "This video demos how to build the latest Microsoft AutoGen Studio with Ollama and Groq on Kaggle Free GPU Step by Step." - @jacobcdev . Open source ftw!.
6
7
68
OpenAI Assistant + AutoGen integration coming very very soon! Stay tuned. .
9
6
67
🚀 New AutoGen release python-v0.5.4 is live — and it's packed with new features and upgrades!. ✨ AgentTool & TeamTool — build multi-agent patterns using tool calls. 🛡️ Azure AI Agent adapter — now available as an Extension. 🐳 DockerJupyterCodeExecutor — sandbox your Jupyter.
7
15
67
📣 New AutoGen release . 0.4.3 is the first release since 0.4.0 with a bunch of new features to try out!. - GraphRAG.- Chat completion model caching.- Semantic Kernel AI Connector adapter.- Laying the foundations for memory (!).- Local Jupyter code execution. Read all about it in.
3
13
64
Excited to announce AutoGen Studio v0.4.2 release updates!. ✅ Improved Gallery with easier UI for editing tools, teams, agents & models.✅ Component validation & testing with real-time error highlighting.✅ Token streaming for better dev experience.✅ Side-by-side session
1
10
63
A very simple API for Magentic-One is now available in the dev13 release of autogen-agentchat 0.4
📢Introducing Magentic-One, a generalist 5-agent multi-agent system for solving open-ended web- and file-based tasks. 🤖🤖🤖🤖🤖. Magentic-One represents a significant step towards agents that can complete tasks that people encounter in their daily lives and can achieve strong
1
14
58
To our collaborators & community:. We’ve seen questions about AutoGen forks/clones vs. the official project. Here's a summary of the latest. Please share with others. - The official repo is - We're actively working on AutoGen v0.2, with v0.4 innovations.
9
24
55
🌻.NET + AutoGen. An initial release of new .NET for AutoGen is out!.- Build event-driven LLM agents in .NET.- Make .NET agents interact with Python agents*.- Build distributed agents*.Support for a high-level API (similar to Python agentchat) is coming soon.
6
16
53
We've significantly improved the current roadmap for #AutoGen. See: Visit these issues to see how you can get involved and contribute. #opensource
7
3
47
🌻 @LangChainAI + AutoGen. The LangChainToolAdapter allows AutoGen agents to access a wide range of LangChain tools, expanding the functionality available for AutoGen workflows.
5
9
49
🌻 pip + AutoGen . Installing microsoft/autogen? Access to 'pyautogen' pip package is blocked for us, please do not use it, instead use:.- autogen-core: Actor model for agents.- autogen-agentchat: API for rapid prototyping.- autogen-ext: 1P extensions.- autogenstudio: GUI for
3
10
48
🌻AutoGen Python v0.4.5 released with a few exciting features:. - Streaming tokens from Agents and Teams.- Native support for R1-style reasoning output.- Partial functions as tools. Adds new samples to demonstrate these new features:. - AgentChat + @chainlit_io with streaming.-
1
9
48
🌻Microsoft Research Forum + AutoGen. Register below to attend the lightning talk on Feb 25th. v0.4 introduces many exciting features based on user feedback, including:.- layered, async, & event driven architecture.- cross language support.- improved scalability and modularity. -.
AutoGen update! On Feb 25, @bansalg_ will introduce a transformative update that builds on user feedback and redefines modularity, stability, and flexibility to empower the next generation of agentic AI research and applications. Register here:
3
10
48
🌻 SOTA agents in your console. MagenticOne (M1) is a 5-agent that can perform a wide variety of file and web related tasks. It achieves good generalization performance on challenging agentic benchmarks. It contains 5 agents:.- WebSurfer.- FileSurfer.- Coder.- Executor.-
2
11
47
📙 AutoGen Tutorials. We are starting to create tutorial videos about AutoGen v0.4! Comment and let us know which topics you'd like us to cover!.
13
8
45
🌻 AutoGen v0.4.6 (python) is out. Adds several features.- Adapter for @AnthropicAI Model-Context Protocol.- Agents can call remotely hosted tools through HTTP.- Magentic-One can now use text-only LLMs.- GroupChat now works better with LLaMa @Meta .- Gemini model client @Google
2
7
46
🌻 AutoGen v0.4.4 is out and adds exciting features:. - Serializable configs for agents & teams.- LLM clients for Azure & GitHub models (Phi-4, Mistral, Cohere).- Prettified console for MagenticOne CLI.- In-memory cache for LLM clients .- Documentation updates, guide for Ollama,.
1
10
41
🚀 Version 0.4.0.dev13 is here!. This release removes previously deprecated features, so ensure your code runs without warnings on dev12 before upgrading. An initial migration guide is available to help: We're nearing the full 0.4.0 release! Your
2
11
40
🌻 OpenAI o3-mini + AutoGen. AutoGen allows creating complex, SOTA agent teams with support for latest thinking models such as o3. In this example, we are powering a five-agent team built using AutoGen (called MagenticOne) with o3-mini. Agents include:.- web surfer.- file surfer
2
6
40
Working on AI Agents and attending #NeurIPS2023??. Don't miss in-person talk by @Chi_Wang_ on @pyautogen . Details:
1
11
39
🌻 Python async API + AutoGen. AutoGen now uses Python's async API, benefiting agentic workflows by enabling:.- Concurrent AI agents and teams.- Non-blocking operations (e.g., LLM calls, tool use).- Background tasks.Below, e.g., multiple GPT-4o agent teams run concurrently and
0
8
40
📢As part of our ongoing work to improve AutoGen, we have been rewriting AutoGen Studio. It is still early but here are some of the new features enabled by the AutoGen 0.4 preview API. - Real-time agent updates streaming to the frontend.- Run control: You can now stop agents
1
10
39
Open models are catching up with GPT-4, and the space of model serving stacks is evolving very quickly. To help you get started with using open models with #AutoGen, we have published documentation on how to use AutoGen with LM Studio, LiteLLM, Ollama, Together AI, and vLLM.
1
7
38
Checkout this new talk about MagenticOne and AutoGen by @HsseinMzannar at @allen_ai @SemanticScholar . Link:
1
14
37
🌻 Reaching 40k stars even faster. Late last year, the major updates and new features we shipped created an inflection in AutoGen’s growth (star history), enabling us to reach the 40k GitHub stars milestone faster. Congrats to the community and especially the
1
7
37
AutoGen's code execution capabilities have gotten an upgrade! You can use a Jupyter kernel to maintain a stateful session for code execution. 🤖🌎🔧. Learn more here:
1
7
36
2
10
35
If you want agents that improve over time, checkout new #AutoGen blogpost on teachable agents by @rickyloynd . "We introduce TeachableAgent . so that users can teach their LLM-based assistants new facts, preferences, and skills.".
2
8
31
Appreciate the support @AndrewYNg 🤩🤖🤖🤖.
I’ve been a fan of @pyautogen as a multiagent programming framework for awhile. It was great hosting two of its leaders @Chi_Wang_ and @qingyun_wu to discuss agent design patterns!
0
2
28
🔔 New feature alert: Workbench 🛠️. A single, consolidated way to give your Autogen agents access to tools that share sessions, resources, and state. Instead of listing every adapter, you pass one Workbench object that keeps a common session across tools. Why this matters?. ▪️
1
10
34
Checkout our new blog post on the first version of #AgentEval — a framework built with #AutoGen to help you gauge how LLM applications add utility for end-users! . We'd love to hear what you think!. Details:
1
3
30
🌻Serving AutoGen Agents. @wronkiew created a way for users to interact with AutoGen 0.4 agents using a familiar chat interface such as @OpenWebUI . Checkout this new open source AutoGen sample below.
I just uploaded my AI agent server to GitHub. Talk to your @pyautogen AutoGen 0.4 agents and teams using @OpenWebUI or any other client using the OpenAI Chat Completion API.
1
8
32
🌻 AutoGen v0.4 release highlights. - Ecosystem contains framework, dev tools, apps.- Layered architecture (Core | AgentChat | Extensions).- Core implements actor model for multi-agents.- AgentChat provides simple, easy to use, async API.- Ext provides advanced clients, tools,.
Announcing AutoGen 0.4, fully reimagined library for building advanced agentic AI systems, developed to improve code quality and robustness. Its asynchronous, event-driven architecture is designed to support dynamic, scalable workflows. Learn more:
1
10
31
v0.2.0 is out! 🚀 . New features include OpenAI structural output support, improved OpenAI integration with OpenAI v2.0, and more. Release note: Get-start samples: cc: @bigmiao6 . #AutoGen #Microsoft
1
9
32
@imeshayat Yes! There are many! Join our Discord channel for more (link in bio). Sample app: Agent Cloud: Gradio: Chainlit:
1
13
18
📢Introducing Magentic-One, a generalist 5-agent multi-agent system for solving open-ended web- and file-based tasks. 🤖🤖🤖🤖🤖. Magentic-One represents a significant step towards agents that can complete tasks that people encounter in their daily lives and can achieve strong
0
6
30
This new blog by @g3rardomz shows how you can use.@e2b_dev with #AutoGen to execute agent generated code in the cloud. Specifically, it shows how you can use it with autogen-core, the new event-driven base layer for the AutoGen stack that is currently in-preview.
The new version 0.4 of @pyautogen was recently released, and it's a major one. Since many of us use agents to generate code, here's a straightforward way to run your agent's code in the cloud using @e2b_dev.
1
2
30
🔔 Important update with AutoGen v0.4, please retweet. As noted in a previous statement from the team, the pyautogen PyPI package is blocked for us. Please use autogen-agentchat, now at v0.4.x, for the latest features and updates. Higher pyautogen versions on PyPI are from a.
4
10
28
This community extension allows AutoGen agents to use tools via @AnthropicAI's MCP. GitHub: richard-gyiko/autogen-ext-mcp
3
4
29
@AndrewYNg @Chi_Wang_ @qingyun_wu Thank you, Andrew!! . Folks, here's the link to #AutoGen OSS framework: More links (research/Discord, etc) in bio!.
2
4
27
Another productive week for #AutoGen with 50 merged PRs and 13 authors 🤖🤖🤖. Checkout:. - *Preview* of v0.4 at - v0.2 at
6
5
28
🌻 @deepseek_ai + @github + AutoGen. - DeepSeek-R1 (@deepseek_ai) is an open-source language model offering advanced text capabilities at a lower cost. - GitHub Models provides free pre-trained AI models for developers to experiment and integrate AI directly within GitHub. -.
Now available in GitHub Models: DeepSeek-R1 🥳 Try it out today.
1
7
27
We’re thrilled to be back at #NeurIPS this year! 🎉 . Don’t miss the **AutoGen Expo Workshop**—details below. 🚀 Come join us for an engaging and exciting session.
0
8
26
🌻 Community Extensions + AutoGen. You can create and publish your own open-source extensions for the AutoGen ecosystem, e.g., to implement advanced clients, tools, and teams. Creating extensions is easy. - name the package "autogen-".- use common interfaces from core and
0
3
26
Checkout new blog post on AutoGen Studio, a low code interface for creating agents!.
AutoGen Studio—built on Microsoft’s flexible open-source AutoGen framework for orchestrating AI agents—provides a user-friendly interface that enables developers to rapidly build, test, customize, & share multi-agent AI solutions, with little or no coding.
0
6
26
Checkout out new #AutoGen short course in collaboration with @DeepLearningAI.
New Agentic AI short course! AI Agentic Design Patterns with AutoGen, taught by @MSFTResearch's @Chi_Wang_ and @penn_state's @qingyun_wu, shows you how to use AutoGen to implement agentic design patterns like multi-agent collaboration, sequential and nested chat, reflection, tool
1
5
24
New #AutoGen blogpost:. You can now control transitions between agents using finite state machines!. See:
0
9
24
🚨 New Blog: What's New in AutoGen?. Five months have passed since our initial spinoff! Checkout out our reflection of lessons, features, and milestones!. See:
0
6
24
🚀🚀AutoGen is trending @github repo of the month again! . A collaborative achievement of many -- our developer community, customers, maintainers, project managers, designers, and researchers. 💐
0
3
24
#AutoGen is enabling both app development AND research!. It is not just allowing developers to create sophisticated applications, it is also enabling novel research on AI agents. Actively maintained, open-source frameworks like AutoGen can especially help conduct reproducible
0
6
24
Checkout this new paper from our team on debugging multi-agent workflows.
Excited to share our #CHI2025 paper: "Interactive Debugging and Steering of Multi-Agent AI Systems"!. As 🤖 AI agent teams 🤖 powered by LLMs take on complex tasks, debugging and steering these agents becomes crucial. How can developers better understand AI agent behavior?
1
3
22
Checkout this tutorial of #AutoGen Magentic-One by @MervinPraison.
Microsoft Magentic One: BEST Generalist AI Agents Framework EXPOSED!. 🤖 Hands-on Tutorial.📚 What is Magnetic One?.🎯 What are the Agent Types?.⚡ How to Install?.💡 What are Example Use Cases?.🔄 How does It Work?. @satyanadella RT Please.@Microsoft @MicrosoftAI @pyautogen
0
5
22
#AutoGen featured in this week's edition of The Batch by @DeepLearningAI and @AndrewYNg!. "What if, instead of having only a single agent, we have one agent to do research and gather information, a second agent to analyze the research, and a third to write the final report?".
2
4
22
. for example, this user guide shows how you can implement handoffs with the autogen-core, and ensure that the implementation remains scalable, flexible, and natively asynchronous!. Checkout the guide: Repo: 2/2
0
0
21
🎉Congratulations to our amazing team members, @ekzhu and @jack_gerrits who are currently top trending developers on GitHub. #AutoGen #OpenSourceAgents
1
1
22
🚀 New AutoGen release python-v0.5.1!. - Support for custom message types. - Enhanced structured output. - New Azure AI Search tool. - Added cancellation support in Code Executors. - Token limit for long-running agents. - Various bug fixes.
1
12
22
Reflect-RL utilizes #AutoGen for a two-player system, optimizing LMs for multi-round interactions with online RL. Dividing tasks between "reflection" and "policy" agents simplifies & improves online RL training: . @VectorZhou @SimonShaoleiDu @beibin79
0
5
21
🌻 Create AI agents with Phi-4 (multimodal) + AutoGen. You can access the latest multi-modal Phi-4 models via @github models and then use it to power your AutoGen agents
1
5
20
DYK you can use AutoGen to solve complex tasks by *decomposing* them in simpler tasks? . While AutoGen supports many ways to do this, see one notebook example below: .
0
1
19
We've been cooking! Another productive week with 42 merged PRs from 19 authors! . To become a contributor, see:. #AutoGen #Agents #OpenSource
1
3
20
Checkout this new course on AutoGen by @amit_aiml.
🚀 @Microsoft has recetly released a rearchitectured version of @pyautogen v0.4 — the future of intelligent, modular multi-agent AI systems!. To help you get hands-on, I’ve launched a new course:. 🎓 Build Intelligent Multi-Agent Applications with AutoGen (v0.4).
0
2
20
AutoGen + Azure OpenAI = ?. Check this series of building AI agent applications using AutoGen and Azure OpenAI, with Python or no code.
0
4
19
🌻 @chainlit_io + AutoGen. Chainlit is an open-source Python package to build chat apps for AI. Here's how you can build a chat app that uses Chainlit for UI and AutoGen for backend. - Define any AutoGen team.- Send user queries from the UI to the team.- Run the team with the
1
5
20
"It is the coolest piece of technology I've seen since ChatGPT".- @MatthewBerman . Checkout the playlist here:
0
2
20
Dive into AutoGen's latest feature: Custom Models! . Now you can decide not only which models are used in AutoGen but also how they get used, for a tailored AI experience. Learn more and get started: ..by @olgavrou. #AutoGen #CustomModels #LocalInference.
1
7
19