<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Multi-Agent Systems on DATATWEETS</title><link>https://datatweets.com/courses/ai-agents/multi-agent-systems/</link><description>Recent content in Multi-Agent Systems on DATATWEETS</description><generator>Hugo</generator><language>en</language><copyright>Copyright (c) 2025 Datatweets</copyright><lastBuildDate>Sun, 28 Jun 2026 09:00:00 +0200</lastBuildDate><atom:link href="https://datatweets.com/courses/ai-agents/multi-agent-systems/index.xml" rel="self" type="application/rss+xml"/><item><title>Lesson 1 - Why and When to Use Multiple Agents</title><link>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-1-why-and-when-multiple-agents/</link><pubDate>Fri, 30 Jan 2026 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-1-why-and-when-multiple-agents/</guid><description>Piling every job onto one agent bloats its prompt and blurs its focus. Splitting work across specialists helps — at the cost of coordination. Learn the real trade-off, why to exhaust a single agent first, and the three multi-agent patterns this module builds: agents-as-tools, routing, and orchestrator-workers.</description></item><item><title>Lesson 2 - Agents as Tools</title><link>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-2-agents-as-tools/</link><pubDate>Fri, 30 Jan 2026 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-2-agents-as-tools/</guid><description>A specialist agent is exposed to a supervisor as a plain tool: the supervisor calls research_destination(question) like any tool, and behind it a full sub-agent runs its own run_agent loop with its own retrieval tool. You build it in stages, watch a retrieved fact propagate from the sub-agent all the way up into the supervisor&amp;rsquo;s final answer, and see why context isolation makes this pattern so clean.</description></item><item><title>Lesson 3 - Routing</title><link>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-3-routing/</link><pubDate>Fri, 30 Jan 2026 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-3-routing/</guid><description>Routing sends each request to exactly one specialist. A lightweight classification call reads the request, picks a label, and dispatches to the matching handler — cheaper and sharper than one agent handling every category. You&amp;rsquo;ll build route(), add a fallback so a bad label never crashes the system, and see it verified against an SDK-shaped mock.</description></item><item><title>Lesson 4 - Orchestrator-Workers</title><link>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-4-orchestrator-workers/</link><pubDate>Fri, 30 Jan 2026 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-4-orchestrator-workers/</guid><description>Routing picks one specialist; orchestrator-workers runs many and combines them. Build an orchestrator that decomposes a goal into subtasks, runs a worker per subtask, and synthesizes one plan. Learn why independent workers can run in parallel, and when a job is really sequential plan-then-execute instead.</description></item><item><title>Lesson 5 - Guided Project: A Multi-Agent Atlas</title><link>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-5-guided-project-a-multi-agent-atlas/</link><pubDate>Fri, 30 Jan 2026 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/multi-agent-systems/lesson-5-guided-project-a-multi-agent-atlas/</guid><description>Put the module together on Atlas. Build a small team — a retrieval-grounded researcher, a budget analyst, and an itinerary writer — then wire in all three patterns: the researcher as a tool, a router that dispatches by category, and an orchestrator that decomposes a trip into subtasks and synthesizes one grounded, cited plan.</description></item></channel></rss>