Module 1 · Architecture · Lesson 1

Multi-agent systems

When splitting work across a lead agent and specialist subagents pays for itself, what it costs in tokens, latency and reliability, and how to build it in Claude Code, the Agent SDK and eve.

75 min 1 video 3 exercises 7-question quiz Answers

Setup for the exercises:Install Claude Code Node.js, pnpm and TypeScript Durable agents with eve Get an OpenRouter key Local models with Ollama

By the end of this lesson you can

Explain the orchestrator-workers architecture and the context-isolation argument for it
Estimate the token cost, latency and end-to-end success rate of a multi-agent design before building it
Classify real tasks as good or poor fits for multi-agent work using the shared-context and dependency tests
Define specialist subagents in Claude Code, the Agent SDK and eve, with scoped tools and structured hand-offs
Name the common multi-agent failure modes and the checks that catch them

A residential purchase arrives as a contract pack: official copies of the title register and plan, the TA6 and TA10 forms, a local authority search, drainage and water and environmental searches, and for a flat a lease and a management pack. A single agent can read all of it, but by the time it reaches page 40 of the lease its context holds 100,000 tokens of everything else. A tempting alternative is a lead agent that hands each document to a specialist subagent and assembles their findings. This lesson is about when that alternative is better, when it is worse, and how to tell before you build it.

Multi-agent systems

When splitting work across a lead agent and specialist subagents pays for itself, what it costs in tokens, latency and reliability, and how to build it in Claude Code, the Agent SDK and eve.

Multi-agent systems

Multi-agent systems

Sign in to keep reading.

Sign in to keep reading.