magentic-ui README

The project's README from github.com/microsoft/magentic-ui. As one markdown file: README.md.

MagenticLite

Big tasks. Small models.

MagenticLite is the next generation of Magentic-UI — an agentic application from Microsoft AI Frontiers, redesigned to do more with less. It pairs an on-device-friendly orchestrator model (MagenticBrain) with a specialized browser-use model (Fara) so you can automate real work without depending on frontier-scale compute.

See it in action

Click any task below to expand and watch MagenticLite handle it end to end.

Fill expense forms

https://github.com/user-attachments/assets/a982b60d-127e-4310-96e6-9f094dffb7f0

Find prices for recipe ingredients

https://github.com/user-attachments/assets/7ff87917-752f-4e40-b319-8036e08ab8f9

Find and book a restaurant

https://github.com/user-attachments/assets/ee5f967e-ee1b-4554-b329-36b56b595aa4

Organize local files

https://github.com/user-attachments/assets/aff385dd-f856-4db4-8f26-08fd0f940aa4

Quick start

These steps get you running on macOS or Windows (WSL). Need more detail or a different platform? See the Installation Guide.

1. Install MagenticLite

# Create a project directory
mkdir magentic-lite && cd magentic-lite

# Create and activate a virtual environment
uv venv --python=3.12 --seed .venv
source .venv/bin/activate

# Install the latest 0.2.x release from PyPI
uv pip install "magentic_ui>=0.2.0"

2. Run

magentic-ui --port 8081

Open http://127.0.0.1:8081/ and follow the in-app onboarding to connect a model endpoint. If you don't have one yet, see the Model Hosting Guide.

Looking for the previous Magentic-UI 0.1 release (optimized to run with frontier models)? It lives on the magentic-ui-0.1 branch.

Documentation

Doc What's in it
Installation Supported platforms, macOS / WSL prerequisites, install + run commands
Build from source How to run MagenticLite from a clone of this repo (uv sync, pnpm dev, etc.)
Model hosting guide End-to-end walkthrough of standing up a model endpoint and connecting MagenticLite to it
Configuration Sandbox, agent mode, tool approval, full config.yaml example
Troubleshooting Common errors and how to fix them
Limitations Tasks and usage patterns MagenticLite doesn't handle well today
Transparency Note Intended uses, responsible-use guidance, risks, and mitigations

License

Microsoft, and any contributors, grant you a license to any code in the repository under the MIT License. See the LICENSE file.

Microsoft, Windows, Microsoft Azure, and/or other Microsoft products and services referenced in the documentation may be either trademarks or registered trademarks of Microsoft in the United States and/or other countries. The licenses for this project do not grant you rights to use any Microsoft names, logos, or trademarks. Microsoft's general trademark guidelines can be found at http://go.microsoft.com/fwlink/?LinkID=254653.

Any use of third-party trademarks or logos are subject to those third-party's policies.

Privacy information can be found at https://go.microsoft.com/fwlink/?LinkId=521839.

Microsoft and any contributors reserve all other rights, whether under their respective copyrights, patents, or trademarks, whether by implication, estoppel, or otherwise.

magentic-ui (microsoft/magentic-ui)

Searching Database…

10.1K ·GitHub
Magentic-ui
What is this?MagenticLite is a free, open-source agent application from Microsoft AI Frontiers that automates real work — web research, form filling, file management — across the browser and your local file system, built to run on small models.
Is it a desktop app or something I use in the browser?It is a locally installed app with a web interface: you run it on your machine and drive the agent from the browser, while it can also work with your local files.
How do I get it running?Two quick-start steps: install MagenticLite, then run it — supported on macOS and Windows (WSL). There is also a build-from-source path (uv sync, pnpm dev).
Do I need a frontier model or a paid API key for this?No — it is designed for small models: MagenticBrain is an on-device-friendly orchestrator and Fara is a specialized browser-use model, so no frontier-scale compute is required. A model hosting guide walks you through connecting an endpoint.
Is there anything it handles badly?Yes. It is experimental, and the docs include a Limitations page listing the tasks and usage patterns MagenticLite does not handle well today.
How is it different from the original Magentic-UI?MagenticLite is the next generation, redesigned to do more with less — it targets small models instead of frontier-scale compute. The previous Magentic-UI 0.1 release, optimized for frontier models, still lives on the magentic-ui-0.1 branch.
What are MagenticBrain and Fara?MagenticBrain is MagenticLite's on-device-friendly orchestrator model, and Fara is a specialized browser-use model. Together they let the agent automate real work without heavy compute.
Can I configure sandboxing and approvals?Yes — the Configuration doc covers sandbox settings, agent mode, and tool approval, including a full config.yaml example.
What are similar projects?collab-agent, hive, fara, open-dots, bytebot, Agent-S, SystemOneHarness, midscene
Which topics is it in?autogen, cua, task-automation, computer-use-agent, file-management, browser-use, wsl, human-in-the-loop, computer-use, microsoft, sandbox, browser-automation
Creator
microsoft
Size
52.4 MB
Stars
10.1K
Updated
CommitJul 23Releasev0.2.1May 21
Age
1y
Commits · 90 Days