Microsoft releases multi-agent Web operating system! Make AI a truly "controllable, collaborative, and transparent" web execution assistant!

Following MAgentic-One, Microsoft launched a new generation of multi-agent Web operating system: Magentic-UI !

A web automation tool driven by a multi-agent system that can:

• • Automatically browse web pages, click, fill out forms, etc.
• • Generate and execute code to analyze and process web page data or files
• • Multi-agent collaboration, with collaborative planning + collaborative execution capabilities
• • Automatically save successful task processes for quick reuse later

Simply put: it is a super AI assistant that can understand web pages + write code + execute operations!

Magentic-UI is different from other browser products in that it has a transparent and controllable interface. AutoGen The framework integrates powerful multi-agent collaboration capabilities, supports users to intervene, pause or take over operations at any time, and can save successful plans for reuse, achieving both efficiency and security.

Key Features

• •  Collaborative Planning : Co-create and approve step-by-step plans through chat and the plan editor.
• •  Collaborative tasks : interrupt and guide task execution directly from a web browser or via chat. Magentic-UI can also request clarification and provide assistance when needed.
• •  Operation protection : Sensitive operations are performed only with explicit user approval.
• •  Plan Learning and Retrieval : Learn from previous runs to improve future task automation and save them to a plan gallery. Automatically or manually retrieve saved plans for future tasks.
• •  Parallel task execution : You can run multiple tasks in parallel, and the session status indicator tells you when your input is needed or the task is completed.
• •  Multi-agent collaboration : web page operation, code generation/execution, file processing, clear division of labor, doubled efficiency.
• •  Applicable to multiple scenarios : supports web data capture, form filling, code analysis, and file processing, suitable for scenarios such as data analysis, automation tasks, and software development.

Quick Start

Magentic-UI is very user-friendly to install and use. The official website provides detailed documentation (GitHub) and supports local and cloud (Azure) deployment.

① Install Python 3.10+ and virtual environment

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

② Install Magentic-UI using pip

pip install magentic-ui

③ Start port service

magentic-ui --port 8081

If you want to use the Azure model or Ollama, install the optional dependencies:

# for Azure
pip install magentic-ui[azure] 
# for Ollama
pip install magentic-ui[ollama]

Of course, it also supports source code construction of the front-end and back-end. The detailed operation process can be viewed in the project Readme document.

Last browser visit http://localhost:8000, the left panel creates the task session, and the right panel displays the schedule and browser operations.

Applicable scenarios

Magentic-UI's transparency and versatility make it suitable for a variety of complex scenarios:

• •  Web data crawling and analysis : Automatically extract data (such as prices, comments) from websites and generate Python code for analysis, suitable for market research or data science.
• •  Automated form filling : Process complex forms (such as booking a restaurant, purchasing a product), with user approval at each step, securely and efficiently.
• •  Code generation and debugging : Extract code snippets (such as C++ functions) from web pages, automatically generate test cases or optimize code to facilitate development.
• •  File processing and reporting : Analyze local files (such as PDF, CSV) and generate Markdown reports, suitable for scientific research or business analysis.
• •  Task automation : Automate multi-step tasks (such as “find flights + generate price chart”) and plan reuse to make repetitive tasks more efficient.

Last words

Magentic-UI uses Microsoft's open source Magentic series as the intelligent agent coordination core, which can browse web pages, operate DOM, execute JS/scripts, and use code analysis tools.

More importantly, it puts you in full control of the entire process: collaborative planning, intervention at any time, and safety protection, with every step being transparent and controllable.

Users can take over tasks in real time (human-machine hybrid collaborative execution), support UI interaction + command-based operations, and are suitable for developers and non-technical personnel.

If you want AI to automatically help you browse the web, click, collect information, write code and analyze data, Magentic-UI is one of the very good open source solutions!