A multi-agent system splits a job across several agents, each with its own instructions, tools and sometimes its own model. A researcher gathers facts, a writer drafts, a reviewer checks. The agents are coordinated by a pattern: a supervisor that hands out tasks, direct handoffs from one agent to the next, or several agents debating until they agree.
More agents are not automatically better. Each one adds model calls, latency and new ways to fail, so the first question is whether a single agent with better tools would do. These guides help you answer that and then build the system when the answer is yes: a working multi-agent setup in Python, the main orchestration patterns and their trade-offs, and reasoning patterns such as debate that use several model roles inside one agent.
The code in these guides runs on Promptise Foundry, an open-source Python framework for AI agents and MCP servers.