Microsoft AutoGen Tutorial: Get Started in 5 Minutes [2026]

What it does:

Agents are defined as Python objects with configurable system prompts, LLM backends, tools, and message-handling logic. The AssistantAgent and UserProxyAgent base classes cover the most common patterns, and developers can subclass them to create specialized roles such as planners, critics, or domain experts.

Use case:

What it does:

The GroupChat and GroupChatManager classes allow multiple agents to participate in a shared conversation, with the manager selecting the next speaker based on rules, round-robin, or LLM-based routing. This enables team dynamics such as brainstorming, debate, and hierarchical review.

Use case:

What it does:

Agents can write and execute Python code in local processes or isolated Docker containers. The framework handles code extraction from LLM outputs, runs it safely, captures stdout/stderr, and returns results to the conversation for iterative refinement.

Use case:

What it does:

UserProxyAgent supports three human input modes — ALWAYS, TERMINATE, and NEVER — letting developers control when a human can intervene, approve actions, or supply missing information during an agent conversation.

Use case:

What it does:

A web-based interface lets users configure agents, skills, and workflows through forms and drag-and-drop, then run them against real LLMs. It is ideal for prototyping, demos, and enabling non-programmers to experiment with multi-agent patterns.

Use case:

What it does:

Agents can be equipped with arbitrary Python functions or OpenAI-compatible tool schemas, letting them call APIs, query databases, invoke external services, and compose results within the conversation loop.

Use case: