Twenty-two chapters, assembled: one repository configured end to end in the order that makes each layer cheap, the question underneath every decision in this handbook, and where the road goes after the CLI.

Four pages cover every failure, and the first job is knowing which one you are on. The routing question, the commands that show what actually loaded, and the handful of location mistakes behind most configuration surprises.

What it actually costs, why an idle session keeps spending, and the security model underneath every permission prompt — including the sentence about what it does not protect you from.

Same engine, different front doors. The one question that sorts every surface, five ways to work away from your terminal, and what your configuration does and does not follow you onto.

Four schedulers and the constraints that pick between them. /loop for polling while you work, routines for work that outlives your laptop, and the payload wrapper that stops a leaked token becoming an instruction.

Four ways to run Claude sessions at once, and the questions that pick between them. Agent view, agent teams and their experimental sharp edges, cross-session messaging, and when parallelism actually pays for itself.

Delegation as context management. What a subagent starts with, what it deliberately does not inherit, when Claude picks one on its own, and the difference between a subagent and a fork.

Three levels of integration, from Claude running git for you to Claude as a CI step. The @claude bot, the two modes a workflow can run in, and what a headless run does with permissions when nobody is there to answer.

When a server earns its place and when Bash already does the job, the native Chrome integration and its safety model, and the prompt shapes that actually get useful work out of both.

The protocol that connects Claude Code to systems it has no tool for. Four transports, three scopes, and the deferral mechanism that stops ten servers from filling your context window before you type.