Claude Code Multiplayer

Your agent finishes. Your teammate's agent finishes. Two agents just independently rewrote the same file. Again.



Meanwhile, teams on Reddit are running six agents in one shared workspace and shipping from it. One of them counted the damage first. 1,879 markdown files. 317 of them written by agents. A 126KB handoff document, 51 days old, that every fresh agent still read as the current state of the project. The difference isn't talent. It's a workspace with an owner on every file.



This book builds the multiplayer workspace that Slack, Y Combinator, OpenClaw and Lindy all shipped inside one month and none of them documented. By day 5, two agents on two machines read the same plan without being told what it is. By day 10, a gate neither agent controls throws back a handoff that claims "done" on work that does not build. By day 20, a broken four-agent run names the step that actually started it instead of the one that finally crashed. By day 30, you have your own number. One agent against the shared workspace, measured on your own work.



The three best-selling books on this shelf all answer the same question, and it isn't this one. They answer how do I run more agents. None of them answers how do we, plural, share one. You can't ask them what wakes the second agent when a message lands. You can't ask them what happens when two agents claim the same file at the same second. You can't ask them how to remove a teammate and prove the agent forgot what they knew.



Every file in this book ships in a free companion repo: thirteen shell gates, one folder per chapter, and fixtures so you can watch each gate refuse and then watch it stay silent. You clone it in chapter 1 and run it in chapter 1.



Here's what you'll build:



1. A shared workspace two agents on two machines both read the same way.

2. A handoff format both agents obey, and the thing that wakes the second agent when a message lands.

3. Proof of what an agent actually did, and a gate outside both agents that throws back a false "done".

4. One owner per file, with the second claim refused by name.

5. A scope for every agent, plus the injection test you re-run after every change to the workspace.

6. A run log that names the first broken step instead of the last one.

7. Your last run priced to the cent, the rate card you build from it, and a rollout pack that onboards a teammate and proves the agent forgets them when they go.



Every week another 150-page book about AI agent teams hits the shelf. One of the titles ranking on this exact search is four pages long. This one runs 190+ pages and builds one workspace end to end. Scroll up and grab your copy.