Report Abuse

Basic Information

This repository is an Azure Samples project that demonstrates how to build an end-to-end multi-agent application using Microsoft Autogen (including Magentic One) together with Azure OpenAI and a React-based UI. It provides a reference architecture and working example that combines a Python backend and a JavaScript frontend to show how multiple agents can be composed, orchestrated, and deployed to Azure. The repo is intended for developers who want a complete sample for testing, running, and deploying advanced multi-agent workflows with event-driven, asynchronous Autogen components. It also surfaces deployment patterns such as one-line azd up via the Azure Developer CLI and shows how to integrate AI Search ingestion, managed identities, secure code execution via containerized sessions, and observability tooling for agent tracing.

Links

Categorization

App Details

Features
The README highlights several core features: a multi-agent framework built on Autogen and Magentic One, a friendly React UI built with Vite, Tailwind and Shadcn, and one-line deployment to Azure using the Azure Developer CLI (azd up). It includes support for Autogen 0.4 and an event-driven asynchronous architecture, secure sandboxed code execution using containerized dynamic sessions, and built-in managed identities to avoid manual credential handling. Observability and debugging features are included, such as tracing of agent interactions and PromptFlow tracing. The project provides optional demo data ingestion into AI Search, local development guidance for backend and frontend, and notes about required tools like Docker, Python and optional UV and Playwright for local execution and browser automation.
Use Cases
This repository helps developers quickly prototype, test, and deploy multi-agent AI applications by providing a ready-made sample with backend and frontend code, deployment instructions, and integration examples. It reduces setup friction with scripted deployment (azd up), recommended environment prerequisites, and optional demo data ingestion for AI Search. Security and operations are addressed through managed identities and sandboxed execution patterns suitable for running untrusted code inside container sessions. Observability and debugging support make it easier to trace agent workflows and diagnose issues. The sample also serves as a learning resource for Autogen 0.4 and Magentic One patterns and demonstrates how to combine Azure OpenAI, AI Search, and a modern React UI for production-like multi-agent scenarios.

Please fill the required fields*