[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-71960":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":16,"forks30d":16,"starsTrendScore":16,"compositeScore":19,"rankGlobal":10,"rankLanguage":10,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":23,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":33,"readmeContent":34,"aiSummary":35,"trendingCount":16,"starSnapshotCount":16,"syncStatus":36,"lastSyncTime":37,"discoverSource":38},71960,"magentic-ui","microsoft\u002Fmagentic-ui","microsoft","MagenticLite is an experimental agent that works across the browser and local file system","https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fresearch\u002Fblog\u002Fmagentic-ui-an-experimental-human-centered-web-agent\u002F",null,"Python",9898,989,70,53,0,22,96,39.99,"MIT License",false,"main",true,[25,26,27,28,29,30,31,32],"agents","ai","ai-ux","autogen","browser-use","computer-use-agent","cua","ui","2026-06-12 02:02:56","\u003Cdiv align=\"center\">\n\u003Cimg src=\"docs\u002Fimg\u002Fmagui-readme-logo.svg\" alt=\"Magentic-UI Logo\">\n\n\n_Automate your web tasks while you stay in control_\n\n[![image](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fmagentic_ui.svg)](https:\u002F\u002Fpypi.python.org\u002Fpypi\u002Fmagentic_ui)\n[![image](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fl\u002Fmagentic_ui.svg)](https:\u002F\u002Fpypi.python.org\u002Fpypi\u002Fmagentic_ui)\n![Python Versions](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.10%20%7C%203.11%20%7C%203.12%20%7C%203.13-blue)\n[![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-2507.22358-b31b1b.svg)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.22358)\n\n\u003C\u002Fdiv>\n\n---\n\nMagentic-UI is a **research prototype** human-centered AI agent that solves complex web and coding tasks that may require monitoring. Unlike other black-box agents, the system reveals its plan before executions, lets you guide its actions, and requests approval for sensitive operations while browsing websites, executing code, and analyzing files.\n*Check out the [demo section](#demos) for inspiration on what tasks you can accomplish.*\n\n## ✨ What's New\n\nMicrosoft latest agentic model [Fara-7B](https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fresearch\u002Fblog\u002Ffara-7b-an-efficient-agentic-model-for-computer-use\u002F) is now integrated in Magentic-UI, read how to launch in \u003Ca href=\"#fara-7b\"> Fara-7B guide\u003C\u002Fa>\n\n\n- **\"Tell me When\"**: Automate monitoring tasks and repeatable workflows that require web or API access that span minutes to days. *Learn more [here](https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fresearch\u002Fblog\u002Ftell-me-when-building-agents-that-can-wait-monitor-and-act\u002F).*\n- **File Upload Support**: Upload any file through the UI for analysis or modification\n- **MCP Agents**: Extend capabilities with your favorite MCP servers\n- **Easier Installation**: We have uploaded our docker containers to GHCR so you no longer need to build any containers! Installation time now is much quicker.\n\n\n## 🚀 Quick Start\n\nHere's how you can get started with Magentic-UI:\n\n```bash\n# 1. Setup environment\npython3 -m venv .venv\nsource .venv\u002Fbin\u002Factivate\npip install magentic-ui --upgrade\n\n# 2. Set your API key\nexport OPENAI_API_KEY=\"your-api-key-here\"\n\n# 3. Launch Magentic-UI\nmagentic-ui --port 8081\n```\n\nThen open \u003Chttp:\u002F\u002Flocalhost:8081> in your browser to interact with Magentic-UI!\n\n> **Prerequisites**: Requires Docker and Python 3.10+. Windows users should use WSL2. See [detailed installation](#️-installation) for more info.\n\n## Alternative Usage Options\n\n**Without Docker** (limited functionality: no code execution):\n```bash\nmagentic-ui --run-without-docker --port 8081\n```\n\n**Command Line Interface**:\n```bash\nmagentic-cli --work-dir PATH\u002FTO\u002FSTORE\u002FDATA\n```\n\n**Custom LLM Clients**:\n```bash\n# Azure\npip install magentic-ui[azure]\n\n# Ollama (local models)\npip install magentic-ui[ollama]\n```\n\nYou can then pass a config file to the `magentic-ui` command (\u003Ca href=\"#model-client-configuration\"> client config\u003C\u002Fa>) or change the model client inside the UI settings.\n\nFor further details on installation please read the   \u003Ca href=\"#️-installation\">🛠️ Installation\u003C\u002Fa> section. For common installation issues and their solutions, please refer to the [troubleshooting document](TROUBLESHOOTING.md). See advanced usage instructions with the command `magentic-ui --help`. \n\n## Quick Navigation:\n\u003Cp align=\"center\">\n  \u003Ca href=\"#demos\">🎬 Demos\u003C\u002Fa>  | \n  \u003Ca href=\"#how-it-works\">🟪 How it Works\u003C\u002Fa>  | \n  \u003Ca href=\"#installation\">🛠️ Installation\u003C\u002Fa>  | \n  \u003Ca href=\"#troubleshooting\">⚠️ Troubleshooting\u003C\u002Fa>  |  \n  \u003Ca href=\"#contributing\">🤝 Contributing\u003C\u002Fa>  | \n  \u003Ca href=\"#license\">📄 License\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n## Demos\n\n\u003Ctable>\n\u003Ctr>\n\u003Ctd width=\"33%\" align=\"center\">\n\n**🍕 Pizza Ordering**  \n*Web automation with human-in-the-loop*\n\n\u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fdc95cf5f-c4b4-4fe0-b708-158ff071e5a9\" width=\"100%\" style=\"max-height: 300px;\">\n\u003C\u002Fvideo>\n\n\u003C\u002Ftd>\n\u003Ctd width=\"33%\" align=\"center\">\n\n**🏠 Airbnb Price Analysis**  \n*MCP agent integration*\n\n\u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fc19ed8c2-e06f-43b7-bee3-5e2ffc4c5e02\" width=\"100%\" style=\"max-height: 300px;\">\n\u003C\u002Fvideo>\n\n\u003C\u002Ftd>\n\u003Ctd width=\"33%\" align=\"center\">\n\n**⭐ Star Monitoring**  \n*Long-running monitoring task*\n\n\u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fd2a463ca-7a94-4414-932d-a69f30fff63b\" width=\"100%\" style=\"max-height: 300px;\">\n\u003C\u002Fvideo>\n\n\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\n\n## How it Works\n\u003Cp align=\"center\">\n  \u003Cimg src=\".\u002Fdocs\u002Fimg\u002Fmagenticui_running.png\" alt=\"Magentic-UI\" height=\"400\">\n\u003C\u002Fp>\n\nMagentic-UI is especially useful for web tasks that require actions on the web (e.g., filling a form, customizing a food order), deep navigation through websites not indexed by search engines (e.g., filtering flights, finding a link from a personal site) or tasks that need web navigation and code execution (e.g., generate a chart from online data).\n\nWhat differentiates Magentic-UI from other browser use offerings is its transparent and controllable interface that allows for efficient human-in-the-loop involvement. Magentic-UI is built using [AutoGen](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fautogen) and provides a platform to study human-agent interaction and experiment with web agents. Key features include:\n\n- 🧑‍🤝‍🧑 **Co-Planning**: Collaboratively create and approve step-by-step plans using chat and the plan editor.\n- 🤝 **Co-Tasking**: Interrupt and guide the task execution using the web browser directly or through chat. Magentic-UI can also ask for clarifications and help when needed.\n- 🛡️ **Action Guards**: Sensitive actions are only executed with explicit user approvals.\n- 🧠 **Plan Learning and Retrieval**: Learn from previous runs to improve future task automation and save them in a plan gallery. Automatically or manually retrieve saved plans in future tasks.\n- 🔀 **Parallel Task Execution**: You can run multiple tasks in parallel and session status indicators will let you know when Magentic-UI needs your input or has completed the task.\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=wOs-5SR8xOc\" target=\"_blank\">\n    \u003Cimg src=\"https:\u002F\u002Fimg.youtube.com\u002Fvi\u002FwOs-5SR8xOc\u002Fmaxresdefault.jpg\" alt=\"Watch the demo video\" width=\"600\"\u002F>\n  \u003C\u002Fa>\n  \u003Cbr>\n  ▶️ \u003Cem> Click to watch a video and learn more about Magentic-UI \u003C\u002Fem>\n\u003C\u002Fdiv>\n\n\n### Autonomous Evaluation\n\nTo evaluate its autonomous capabilities, Magentic-UI has been tested against several benchmarks when running with o4-mini: [GAIA](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fgaia-benchmark\u002FGAIA) test set (42.52%), which assesses general AI assistants across reasoning, tool use, and web interaction tasks ; [AssistantBench](https:\u002F\u002Fhuggingface.co\u002FAssistantBench) test set (27.60%), focusing on realistic, time-consuming web tasks; [WebVoyager](https:\u002F\u002Fgithub.com\u002FMinorJerry\u002FWebVoyager) (82.2%), measuring end-to-end web navigation in real-world scenarios; and [WebGames](https:\u002F\u002Fwebgames.convergence.ai\u002F) (45.5%), evaluating general-purpose web-browsing agents through interactive challenges.\nTo reproduce these experimental results, please see the following [instructions](experiments\u002Feval\u002FREADME.md).\n\n\n\nIf you're interested in reading more checkout our [technical report](https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fresearch\u002Fwp-content\u002Fuploads\u002F2025\u002F07\u002Fmagentic-ui-report.pdf) and [blog post](https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fresearch\u002Fblog\u002Fmagentic-ui-an-experimental-human-centered-web-agent\u002F).\n\n\n## Installation\n### Pre-Requisites\n\n**Note**: If you're using Windows, we highly recommend using [WSL2](https:\u002F\u002Fdocs.microsoft.com\u002Fen-us\u002Fwindows\u002Fwsl\u002Finstall) (Windows Subsystem for Linux).\n\n1. If running on **Windows** or **Mac** you should use [Docker Desktop](https:\u002F\u002Fwww.docker.com\u002Fproducts\u002Fdocker-desktop\u002F) or if inside WSL2 you can install Docker directly inside WSL [docker in WSL2 guide](https:\u002F\u002Fgist.github.com\u002Fdehsilvadeveloper\u002Fc3bdf0f4cdcc5c177e2fe9be671820c7). If running on **Linux**, you should use [Docker Engine](https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002F). \n\nIf using Docker Desktop, make sure it is set up to use WSL2:\n    - Go to Settings > Resources > WSL Integration\n    - Enable integration with your development distro You can find more detailed instructions about this step [here](https:\u002F\u002Fdocs.microsoft.com\u002Fen-us\u002Fwindows\u002Fwsl\u002Ftutorials\u002Fwsl-containers).\n\n\n\n2. During the Installation step, you will need to set up your `OPENAI_API_KEY`. To use other models, review the [Model Client Configuration](#model-client-configuration) section below.\n\n3. You need at least [Python 3.10](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002F) installed.\n\n\nIf you are on Windows, we recommend to run Magentic-UI inside [WSL2](https:\u002F\u002Fdocs.microsoft.com\u002Fen-us\u002Fwindows\u002Fwsl\u002Finstall) (Windows Subsystem for Linux) for correct Docker and file path compatibility.\n\n\n\n### PyPI Installation\n\nMagentic-UI is available on PyPI. We recommend using a virtual environment to avoid conflicts with other packages.\n\n```bash\npython3 -m venv .venv\nsource .venv\u002Fbin\u002Factivate\npip install magentic-ui\n```\n\nAlternatively, if you use [`uv`](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F) for dependency management, you can install Magentic-UI with:\n\n```bash\nuv venv --python=3.12 .venv\n. .venv\u002Fbin\u002Factivate\nuv pip install magentic-ui\n```\n\n\n### Running Magentic-UI\n\nTo run Magentic-UI, make sure that Docker is running, then run the following command:\n\n```bash\nmagentic-ui --port 8081\n```\n\n>**Note**: Running this command for the first time will pull two docker images required for the Magentic-UI agents. If you encounter problems, you can build them directly with the following command:\n```bash\ncd docker\nsh build-all.sh\n```\n\nIf you face issues with Docker, please refer to the [TROUBLESHOOTING.md](TROUBLESHOOTING.md) document.\n\nOnce the server is running, you can access the UI at \u003Chttp:\u002F\u002Flocalhost:8081>.\n\n\n\n### Fara-7B\n\n1) First install magentic-ui with the fara extras:\n\n```bash\npython3 -m venv .venv\nsource .venv\u002Fbin\u002Factivate\npip install magentic-ui[fara]\n```\n\n2) In a seperate process, serve the Fara-7B model using vLLM:\n\n```bash\nvllm serve \"microsoft\u002FFara-7B\" --port 5000 --dtype auto \n```\n\n3) First create a `fara_config.yaml` file with the following content:\n\n```yaml\nmodel_config_local_surfer: &client_surfer\n  provider: OpenAIChatCompletionClient\n  config:\n    model: \"microsoft\u002FFara-7B\"\n    base_url: http:\u002F\u002Flocalhost:5000\u002Fv1\n    api_key: not-needed\n    model_info:\n      vision: true\n      function_calling: true\n      json_output: false\n      family: \"unknown\" \n      structured_output: false\n      multiple_system_messages: false\n\norchestrator_client: *client_surfer\ncoder_client: *client_surfer\nweb_surfer_client: *client_surfer\nfile_surfer_client: *client_surfer\naction_guard_client: *client_surfer\nmodel_client: *client_surfer\n```\nNote: if you are hosting vLLM on a different port or host, change the `base_url` accordingly.\n\n\nThen launch Magentic-UI with the fara agent:\n\n```bash\nmagentic-ui --fara --port 8081 --config fara_config.yaml \n```\n\nFinally, navigate to \u003Chttp:\u002F\u002Flocalhost:8081> to access the interface!\n\n### Configuration\n\n#### Model Client Configuration\n\nIf you want to use a different OpenAI key, or if you want to configure use with Azure OpenAI or Ollama, you can do so inside the UI by navigating to settings (top right icon) and changing model configuration. Another option is to pass a yaml config file when you start Magentic-UI which will override any settings in the UI:\n\n```bash\nmagentic-ui --port 8081 --config config.yaml\n```\n\nWhere the `config.yaml` should look as follows with an AutoGen model client configuration:\n\n```yaml\ngpt4o_client: &gpt4o_client\n    provider: OpenAIChatCompletionClient\n    config:\n      model: gpt-4o-2024-08-06\n      api_key: null\n      base_url: null\n      max_retries: 5\n\norchestrator_client: *gpt4o_client\ncoder_client: *gpt4o_client\nweb_surfer_client: *gpt4o_client\nfile_surfer_client: *gpt4o_client\naction_guard_client: *gpt4o_client\nplan_learning_client: *gpt4o_client\n```\nYou can change the client for each of the agents using the config file and use AzureOpenAI (`AzureOpenAIChatCompletionClient`), Ollama and other clients.\n\n#### MCP Server Configuration\n\nYou can also extend Magentic-UI's capabilities by adding custom \"McpAgents\" to the multi-agent team. Each McpAgent can have access to one or more MCP Servers. You can specify these agents via the `mcp_agent_configs` parameter in your `config.yaml`.\n\nFor example, here's an agent called \"airbnb_surfer\" that has access to the OpenBnb MCP Server running locally via Stdio.\n\n```yaml\nmcp_agent_configs:\n  - name: airbnb_surfer\n    description: \"The airbnb_surfer has direct access to AirBnB.\"\n    model_client: \n      provider: OpenAIChatCompletionClient\n      config:\n        model: gpt-4.1-2025-04-14\n      max_retries: 10\n    system_message: |-\n      You are AirBnb Surfer, a helpful digital assistant that can help users acces AirBnB.\n\n      You have access to a suite of tools provided by the AirBnB API. Use those tools to satisfy the users requests.\n    reflect_on_tool_use: false\n    mcp_servers:\n      - server_name: AirBnB\n        server_params:\n          type: StdioServerParams\n          command: npx\n          args:\n            - -y\n            - \"@openbnb\u002Fmcp-server-airbnb\"\n            - --ignore-robots-txt\n```\n\nUnder the hood, each `McpAgent` is just a `autogen_agentchat.agents.AssistantAgent` with the set of MCP Servers exposed as an `AggregateMcpWorkbench` which is simply a named collection of `autogen_ext.tools.mcp.McpWorkbench` objects (one per MCP Server).\n\nCurrently the supported MCP Server types are `autogen_ext.tools.mcp.StdioServerParams` and `autogen_ext.tools.mcp.SseServerParams`.\n\n### Building Magentic-UI from source\n\nThis step is primarily for users seeking to make modifications to the code, are having trouble with the pypi installation or want the latest code before a pypi version release.\n\n#### 1. Make sure the above prerequisites are installed, and that Docker is running.\n\n#### 2. Clone the repository to your local machine:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fmagentic-ui.git\ncd magentic-ui\n```\n\n#### 3. Install Magentic-UI's dependencies with uv or your favorite package manager:\n\n```bash\n# install uv through https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F\nuv venv --python=3.12 .venv\nuv sync --all-extras\nsource .venv\u002Fbin\u002Factivate\n```\n\n#### 4. Build the frontend:\n\nFirst make sure to install node:\n\n```bash\n# install nvm to install node\ncurl -o- https:\u002F\u002Fraw.githubusercontent.com\u002Fnvm-sh\u002Fnvm\u002Fv0.40.1\u002Finstall.sh | bash\nnvm install node\n```\n\nThen install the frontend:\n\n```bash\ncd frontend\nnpm install -g gatsby-cli\nnpm install --global yarn\nyarn install\nyarn build\n```\n\n#### 5. Run Magentic-UI, as usual.\n\n```bash\nmagentic-ui --port 8081\n```\n\n\n#### Running the UI from source\n\nIf you are making changes to the source code of the UI, you can run the frontend in development mode so that it will automatically update when you make changes for faster development.\n\n1. Open a separate terminal and change directory to the frontend\n\n```bash\ncd frontend\n```\n\n2. Create a `.env.development` file.\n\n```bash\ncp .env.default .env.development\n```\n\n3. Launch frontend server\n\n```bash\nnpm run start\n```\n\n4. Then run the UI:\n\n```bash\nmagentic-ui --port 8081\n```\n\nThe frontend from source will be available at \u003Chttp:\u002F\u002Flocalhost:8000>, and the compiled frontend will be available at \u003Chttp:\u002F\u002Flocalhost:8081>.\n\n\n\n\n## Troubleshooting\n\n\nIf you were unable to get Magentic-UI running, do not worry! The first step is to make sure you have followed the steps outlined above, particularly with the [pre-requisites](#pre-requisites).\n\nFor common issues and their solutions, please refer to the [TROUBLESHOOTING.md](TROUBLESHOOTING.md) file in this repository. If you do not see your problem there, please open a `GitHub Issue`. \n\n## Contributing\n\nThis project welcomes contributions and suggestions. For information about contributing to Magentic-UI, please see our [CONTRIBUTING.md](CONTRIBUTING.md) guide, which includes current issues to be resolved and other forms of contributing.\n\nThis project has adopted the [Microsoft Open Source Code of Conduct](https:\u002F\u002Fopensource.microsoft.com\u002Fcodeofconduct\u002F). For more information, see the [Code of Conduct FAQ](https:\u002F\u002Fopensource.microsoft.com\u002Fcodeofconduct\u002Ffaq\u002F) or contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with any additional questions or comments.\n\n\n## Citation\n\nPlease cite our paper if you use our work in your research:\n\n```\n@article{mozannar2025magentic,\n  title={Magentic-UI: Towards Human-in-the-loop Agentic Systems},\n  author={Mozannar, Hussein and Bansal, Gagan and Tan, Cheng and Fourney, Adam and Dibia, Victor and Chen, Jingya and Gerrits, Jack and Payne, Tyler and Maldaner, Matheus Kunzler and Grunde-McLaughlin, Madeleine and others},\n  journal={arXiv preprint arXiv:2507.22358},\n  year={2025}\n}\n```\n\n## License\n\nMicrosoft, and any contributors, grant you a license to any code in the repository under the [MIT License](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT). See the [LICENSE](LICENSE) file.\n\nMicrosoft, Windows, Microsoft Azure, and\u002For other Microsoft products and services referenced in the documentation\nmay be either trademarks or registered trademarks of Microsoft in the United States and\u002For other countries.\nThe licenses for this project do not grant you rights to use any Microsoft names, logos, or trademarks.\nMicrosoft's general trademark guidelines can be found at \u003Chttp:\u002F\u002Fgo.microsoft.com\u002Ffwlink\u002F?LinkID=254653>.\n\nAny use of third-party trademarks or logos are subject to those third-party's policies.\n\nPrivacy information can be found at \u003Chttps:\u002F\u002Fgo.microsoft.com\u002Ffwlink\u002F?LinkId=521839>\n\nMicrosoft and any contributors reserve all other rights, whether under their respective copyrights, patents, or trademarks, whether by implication, estoppel, or otherwise.\n\n","Magentic-UI 是一个以人为中心的AI代理研究原型，旨在自动化复杂的网页和编码任务。其核心功能包括在执行任务前展示计划、允许用户引导行动以及请求批准敏感操作，从而确保用户始终掌控过程。该项目使用Python开发，并集成了微软最新的Fara-7B模型，支持文件上传分析、MCP代理扩展等功能，且安装简便。适合需要对网页浏览、代码执行及文件分析等任务进行监控与控制的场景，尤其适用于那些希望保持透明度和控制权的技术人员或研究人员。",2,"2026-06-11 03:39:41","high_star"]