Microsoft Releases Magentic-UI, an Open-Source Agentic Web Interface That Can Perform Tasks on the Web

Microsoft launched an agentic internet interface on Monday that may carry out complicated duties on the internet. Announced at the firm’s Build 2025 convention, the new instrument is known as Magentic-UI, and the firm is pitching it as a “human-centred web agent.” The Redmond-based tech large claims that, not like different laptop use brokers, this one collaborates with the consumer to finish a mission and transparently reveals the actions it takes in a separate panel. Since Magentic-UI is an open-source agent, the firm has made it accessible on GitHub in addition to its Azure AI Foundry Labs.

Magentic-UI Arrives as Microsoft’s First Computer Use Agent

In its analysis weblog, the tech large introduced the launch of the Magentic-UI agentic internet interface and detailed its workings. Essentially, it’s a computer-focused community of brokers, much like OpenAI’s Operator or Anthropic’s Computer Use. These brokers can browse and navigate by way of a digital machine and entry apps resembling internet browsers to execute actions. These can be utilized to carry out duties resembling reserving a desk at a restaurant or buying a product on-line.

The tech large states that Magentic-UI can browse the internet, write and execute Python and shell code, and analyse recordsdata. Highlighting the “human-centric” moniker, Microsoft mentioned that the AI agent permits customers to switch its plans, lets customers pause the system and provides suggestions in pure language, and even take management of the browser again.

Magentic-UI’s interface
Photo Credit: Microsoft

Additionally, the agentic community will search consumer approval earlier than taking any motion that’s deemed irreversible. Such actions embody inserting an order from a restaurant, including merchandise to the cart, and extra. Users may modify the command right now. Further, Magentic-UI may study from and keep in mind earlier duties and use that info to finish future duties.

Coming to the structure, it’s constructed on the firm’s Magentic-One agentic system. Microsoft says Magentic-UI is a mix of 4 completely different AI brokers, working collectively to finish a activity. The interface contains an Orchestrator or the lead agent that makes the activity plan alongside the consumer. It additionally incorporates suggestions and delegates sub-tasks to different brokers. Then there’s a WebSurfer agent that’s geared up with an online browser. It can click on, sort, scroll, and go to pages in a number of rounds to finish a activity.

Then there’s a Coder agent that comes with a Docker code-execution container. It can write and execute Python and shell instructions. And lastly, the FireSurfer can be geared up with a Docker code-execution container and file conversion instruments.

Those considering downloading and utilizing the AI agent in an area setting can entry it through GitHub. Microsoft is providing the analysis preview of the agent with a permissive MIT licence. For those that favor to entry the cloud-based model of Magentic-UI, the firm is internet hosting it on its Azure AI Foundry Labs.