# microsoft/magentic-ui

MagenticLite is an experimental agent that works across the browser and local file system

Repository: https://github.com/microsoft/magentic-ui
Canonical: https://ross.abutalabs.com/products/magentic-ui
Homepage: https://www.microsoft.com/en-us/research/blog/magenticlite-magenticbrain-fara1-5-an-agentic-experience-optimized-for-small-models/
Language: Python
License: MIT
License Family: permissive
Topics: agents, ai, ai-ux, autogen, browser-use, computer-use-agent, ui, cua, fara, magenticbrain, magenticlite
Last push: 2026-08-26T23:47:22+00:00

## Health v2 (maintenance only)
Score: 81/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 84, longevity 34
- inputs: {"age_days": 485, "days_push": 7, "days_rel": 104, "gap_med": 26, "n_releases_24m": 10}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 10077, forks 1015 (observed 2026-08-28T04:10:39.488155+00:00)

## What it is
MagenticLite is an experimental agentic application from Microsoft AI Frontiers that automates tasks across the browser and local file system using small, efficient models (MagenticBrain orchestrator plus the Fara browser-use model). It provides a human-centered web UI where users can steer, approve, or take over agent actions, with browser sessions isolated in a sandboxed VM.

## Use cases
- automate filling out web forms like expense reports
- do web research and price comparison with an AI agent
- have an AI agent organize files on my computer
- run a browser automation agent with small local models
- book a restaurant or complete multi-step web tasks automatically
- supervise an AI agent with human approval before critical actions

## When to choose
- you want agentic browser and file automation without frontier-scale models
- you need human-in-the-loop control and approval before critical actions
- you want sandboxed browser sessions for safety
- you're experimenting with multi-agent web automation research prototypes

## When to avoid
- you need a production-hardened, stable agent platform
- you require frontier-model-level performance on complex tasks
- you need a headless automation tool without a UI
- your platform isn't macOS, Windows (WSL), or Linux

## Facets
- artifact type: application
- maturity: experimental
- function: agent-framework, web-scraping, chat-interface, workflow-automation
- domain: artificial-intelligence, web-development
- platform: windows, python
- tags: browser-use, computer-use-agent, small-models, human-in-the-loop, sandboxed-browser, multi-agent, local-file-system, microsoft-research, ai-agents, automation, macos, linux, web-server, docker

## Member repositories
- microsoft/magentic-ui (main) score 81

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:39.488155+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:19:51.544570+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/magentic-ui (fetched 2026-08-28T04:10:39.488155+00:00, sha 56bf597aeb4e)
  - homepage: https://www.microsoft.com/en-us/research/blog/magenticlite-magenticbrain-fara1-5-an-agentic-experience-optimized-for-small-models/ (fetched 2026-08-29T08:19:43.818919+00:00, sha 1e0c5c523ac6)
  - site_page: https://www.microsoft.com/en-us/research/about-microsoft-research (fetched 2026-08-29T08:19:43.828540+00:00, sha 662089aedfa3)
  - site_page: https://learn.microsoft.com/docs (fetched 2026-08-29T08:19:43.830378+00:00, sha f29800be2b9a)
  - site_page: https://www.microsoft.com/about (fetched 2026-08-29T08:19:43.831949+00:00, sha 4435a23e63d2)
  - registry_pypi: https://pypi.org/pypi/magentic-ui/json (fetched 2026-08-29T08:19:43.835124+00:00, sha 834f04a47bdb)
  - site_page: https://choice.microsoft.com (fetched 2026-08-29T08:19:43.833633+00:00, sha 2db73202bb74)
- Data as of 2026-08-30T08:39:29.467469+00:00.