“One agent works. I need five running at once without colliding.”
A lead assigning bounded work to cheap workers, a fan-out past one context window, a bad-handoff gate, and a spend cap that kills a runaway run
A lead assigning bounded work to cheap workers.
Read a free guide from this book first How to Build a Multi-Agent MCP Pipeline in Python · 10 min ▸
Twelve working graph systems on your own repo, a real cost number per run, and one packaged as a client deliverable
Graph engineering: twelve parallel systems on your own repo.
Read a free guide from this book first Give Your Claude Code Subagents Memory Without a Database ▸
Five agents on one queue that resume after a hard kill, never collide on a repo, and stop before the bill lands
Five agents on one queue that survive a hard kill.
A split-stack system where Fable plans and specialist models execute, plus a working /architect router you own
A model team: one architect model planning, specialists executing.
Read a free guide from this book first How to Run a Claude Repo Audit Your Clients Would Pay a Consultant to Do ▸
Ship an MCP server, sell it for $3K-$5K
The tool and server layer, once the fleet needs to reach real systems.
Read a free guide from this book first How to Build Your First MCP Server in Python · 9 min ▸