Claude Code · Knowledge Base · 35 topics
Claude Code Field Guide
A living knowledge base of Claude Code concepts — explained, fact-checked, and kept current. Started from a single explainer video, then checked concept-by-concept against Anthropic's official documentation so you know what still holds up, what's changed, and what's brand new.
Foundations
Start hereWhat Claude Code actually is, and the basic vocabulary you need before anything else in this library makes sense.
What Is Claude Code
A chatbot answers questions in a window. Claude Code takes actions on your actual computer.
The Terminal
The black screen with white text is just another way to talk to your computer — and Claude Code hides almost all of it from you.
Prompts
What you type to tell Claude Code what to do — plain English, no special syntax, and specificity is the whole game.
Tool Use
Claude Code can act because it has tools — Read, Write, Bash, and more — and it picks the right one automatically.
The Context Window
Claude's short-term memory for the current conversation — everything it can “see” right now, and it has a hard limit.
Conversation History & Sessions
Every session is saved automatically. “claude --resume” drops you back into any past conversation, with full context.
Token Usage & Cost Basics
Claude Code is billed by tokens (roughly ¾ of a word each). What you send and receive both count.
Permissions & Safety
Stay in controlHow Claude Code decides what it's allowed to do on your machine — and how you dial that up or down.
Permissions & settings.json
Claude Code can create, edit and delete files, so by default it asks before doing anything significant — and you can pre-approve the safe stuff.
Permission Modes (New Since the Video)
Claude Code now ships several named permission modes — and on paid plans, the default is no longer “ask for everything.”
Denying Access to Files
Not every file should be visible to Claude Code — deny rules make secrets and irrelevant files invisible, not just off-limits.
Sandboxing (New Since the Video)
Instead of approving commands one by one, you can let the operating system enforce a boundary — which files and network domains Claude is allowed to touch.
Flags
Launch-time options that configure a session before it even starts — which model, which tools, how verbose, how permissive.
Making It Personal
Teach it your way of workingGiving Claude Code your preferences, project rules and working style so it stops guessing and starts matching how you work.
CLAUDE.md
The instruction manual Claude Code reads at the start of every session — your rules, conventions and project structure, written once.
Auto Memory
Claude Code notices your patterns and preferences and remembers them across sessions on its own — you can review and edit what it stored.
Compacting Context
When the context window fills up, Claude Code (or you) can summarize the important parts and clear out the noise.
Models & Effort Levels
“Claude Code” refers to a family of models with different speed, intelligence and cost — and you can switch mid-conversation.
Extended Thinking
Claude's ability to reason through a hard problem step by step before acting — on by default, no magic words required.
Power Features
Beyond a chat windowThe features that turn Claude Code from a conversational assistant into a real automation and extension platform.
Slash Commands
Type / and a command name to trigger a repeatable action — /init, /compact, /clear, /help, and any custom one you define.
Skills
Pre-written expert playbooks Claude reads before it builds — install one, or write your own for the tasks you do repeatedly.
Hooks
Custom scripts that run automatically at specific moments — guardrails that don't spend AI tokens or need supervision.
Plugins (New Since the Video)
A single installable bundle of skills, agents, hooks and MCP servers — the packaging layer that didn't exist yet when the video was made.
MCP Servers
Model Context Protocol connects Claude Code to the tools your business actually lives in — Notion, Airtable, Slack, databases, and more.
Multimodal Support
Paste a screenshot instead of describing the bug in words — works in reverse too, for matching a design you like.
Checkpoints & Rewind
Claude Code automatically snapshots your code before every edit — /rewind lets you undo to any earlier point, with fine-grained control.
Parallel & Multi-Agent Work
More than one Claude at onceRunning several Claude Code sessions or agents at the same time, without them overwriting each other's work.
Sub Agents
Specialists that run in their own context window, do focused work, and report back a summary — keeping your main conversation clean.
Agent Teams
Named teammates that talk to each other directly and share a task list — solving the bottleneck of everything routing through one main agent.
Worktrees
Separate working directories that share one repository history, so multiple Claudes can work on different things at once without colliding.
Cross-Session Messaging (New Since the Video)
Let the separate Claude sessions you're running yourself — in different worktrees or terminals — pass messages to each other.
Automation & Real-World Use
Take it out of the chat windowVersion control, unattended runs, scheduling, cost, and where Claude Code fits into a real workflow or business.
Git Integration
Claude Code works directly with git — staging, committing, branching, and opening pull requests — and it powers the checkpoint system underneath.
Headless / Non-Interactive Mode
Run Claude Code with no back-and-forth and no approval prompts, using the -p flag — turning it into a scriptable, pipeable tool.
Scheduled Tasks: /loop, Routines & the “Ralph Loop”
Instead of running once and stopping, Claude Code can keep re-running a prompt until a task is actually done — now built in, not a community trick.
GitHub Actions, GitLab CI & Code Review (New Since the Video)
Claude Code plugged directly into your CI pipeline — automatic PR review and issue triage without anyone manually invoking it.
Working From Anywhere: Desktop, Web & Mobile (New Since the Video)
The terminal is now just one of several surfaces — desktop app, browser, and phone all run the same underlying Claude Code engine.
The Agent SDK (New Since the Video)
The same engine that powers Claude Code, available as a Python/TypeScript library for building your own fully custom agents.
Pricing & Plans
Subscription (Pro/Max) vs. pay-as-you-go API — the video's guidance to prefer a flat monthly plan for predictability still holds.