A development agent whose project context, extensions, and enforceable tool permissions need separate understanding.
Coding & development · Documented · Documentation checked 2026-10-04
Claude Code in terminals and supported development surfaces; Claude chat is not a local shell.
Authentication, installed version, surface, subscription or API billing, and organization policy matter.
What the platform is for
Claude Code works with a repository and development tools to inspect problems, edit files, and verify selected behavior. Its educational starting point is a small reproducible bug. Learn how context is supplied and how changes are reviewed before adding complicated agent orchestration.
Know the product boundary
Claude Code is not the same as Claude chat or Cowork. Authentication, surface, installed version, subscription / API billing, and organization controls matter. Extensions can offer skills, MCP, hooks, or subagents, but those components have different behavior and side effects.
How the components connect
CLAUDE.md provides project context. Skills provide reusable procedures. MCP exposes tools; plugins package extensions. Subagents delegate work, while hooks respond to lifecycle events. Permissions enforce allowed operations outside the prompt. Review and tests connect the produced change to actual expected behavior.
A beginner starting point
Begin with an explanation-only repo task, then a small fix. Preserve existing edits and inspect commands, output, and diff. Add project guidance next. Introduce one trusted read-only extension before attempting hooks, delegated work, or broader service access.
Limits and data choices
Agents can misread context, execute a bad plan, or report checks that need scrutiny. Parallel tasks can conflict over shared resources. A policy written in CLAUDE.md does not enforce permissions. This edition does not exhaustively teach cloud deployment, every SDK, advanced orchestration, or every enterprise configuration.
The technical path
Follow official Claude Code documentation for installation, configuration, hooks, MCP, skills, and managed policy. For API products and application deployment, use Claude Platform documentation under its own authentication and billing. Keep production authority separate from a learning repo.
Feature map
Local repository work and review
Give a coding agent a reproducible problem and review its actual edits and test output.
CLAUDE.md and skills
Keep project context distinct from a reusable task procedure loaded when needed.
MCP, plugins, and tool trust
Add narrowly scoped external capabilities and inspect the extension before loading it.
Subagents and hooks
Separate delegated reasoning from lifecycle-triggered automation and shared side effects.
Permission modes and stopping rules
Use enforceable tool policies rather than relying on prose instructions as access controls.
Complete workflows
A local change with explicit permission review
Implement a small sample fix and verify its behavior.
A read-only extension-assisted review
Use approved issue context and delegated review without external side effects.
Coverage and availability
Models, modes & conversation
Covered: Repository task and permission modes; model availability depends on account.
Search, research & sources
Covered: External retrieval through supported tools; not Claude Research UI parity.
Files, analysis & code
Covered: Code files, tools, tests, and actual command evidence.
Images, video, audio & voice
Covered: Visual context may be surface-specific; consumer media creation is a separate path.
Projects, instructions & memory
Covered: CLAUDE.md and skills; deliberate guidance emphasized.
Editing surfaces & deliverables
Covered: Repository diffs and outputs; current extension capabilities differ.
Agents, skills, plugins & MCP
Covered: MCP, plugins, subagents, and hooks with separate side effects.
Orchestration & scheduling
Covered: Hooks / delegated work explained; full dynamic orchestration not covered.
APIs, SDKs, deployment & evaluation
Separate product: Technical docs linked; API applications / deployment are separate implementations.
Sharing, privacy & governance
Covered: Permission enforcement distinguished from project prose; external scopes reviewed.
Plans, devices & accessibility
Not established: Authentication and surface conditions included; exact budgets and accessibility not established.
Sources and verification
Documentation checked 2026-10-04. Examples are illustrative unless described otherwise.
Claude Code overview · Anthropic
Configure permissions · Anthropic
Features and extension relationships · Anthropic
Skills: instructions and resources · Anthropic
Use plugins in Claude · Anthropic