How I Built Anthropic's Multi Agent Researcher (in Python)

Description

I walk through a practical implementation of a multi-agent research system using the Claude Agents SDK, closely inspired by Anthropic’s own researcher architecture. This video focuses on how the system is actually wired together: defining sub-agents, introducing a lead agent, coordinating queries, and executing the full research loop end-to-end. This is intended for developers and technically curious practitioners who want to understand why multi-agent systems are structured the way they are, and what you gain by implementing them in code. 00:00 Introduction 00:40 Setting up the Environment 01:21 Claude Agents SDK 02:20 Defining the Sub Agents 03:15 Why Code? 03:45 Defining the Lead Agents 05:02 The Query Loop 05:46 Running the Agent 06:13 Final Results 06:56 Resources and Remarks Resources: Anthropic's Engineering Blog Post: https://www.anthropic.com/engineering/multi-agent-research-system Github Repository: https://github.com/bronsonhill/anthropic-multi-agent-system My goal is to share new AI engineering information clearly. Let me know how I did!