Glossary
Glossary
Section titled “Glossary”Definitions for terms used throughout this guide. For model-level concepts (tokens, temperature, RAG), see the Anthropic platform glossary.
Agent teams
Section titled “Agent teams”Multiple independent Claude Code sessions coordinated by a team lead, with a shared task list and peer-to-peer messaging. Unlike subagents, which run within a single session, teammates each have their own context window and you can interact with any of them directly. This is an experimental feature; it requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 to enable. See §4 Agents.
Agentic loop
Section titled “Agentic loop”The cycle Claude works through for every task: gather context, take action, verify results, repeat until done. Each tool use returns information that informs the next step. Most extension points (hooks, skills, MCP) plug into specific phases of this loop. See §1 How Claude Code Works.
Auto memory
Section titled “Auto memory”Notes Claude writes for itself based on your corrections and preferences, stored per git repository under ~/.claude/projects/. The first 200 lines or 25 KB of MEMORY.md loads at session start. All worktrees of the same repository share one auto memory directory. See §3 Memory.
Auto mode
Section titled “Auto mode”A permission mode where a separate classifier model reviews each action in the background instead of showing you approval prompts. The classifier blocks scope escalation, untrusted infrastructure, and prompt injection attempts. It never sees tool results, making it immune to injected instructions. This is a research preview. See §3 Permission Modes.
Bare mode
Section titled “Bare mode”A startup flag (--bare) that skips auto-discovery of hooks, skills, plugins, MCP servers, auto memory, and CLAUDE.md. Only flags you pass explicitly take effect. Recommended for CI and scripted calls where you need identical behavior across machines. Available since v2.1.81. See §9 Headless/Non-interactive Mode.
Bundled skills
Section titled “Bundled skills”Prompt-based playbooks included with Claude Code, such as /batch, /code-review, /debug, and /loop. Unlike built-in commands (which execute fixed logic), bundled skills give Claude a detailed prompt and let it orchestrate the work. They can spawn agents, read files, and adapt to your codebase. See §5 Skills.
Channel
Section titled “Channel”An MCP server that pushes events into your running session so Claude can react to things that happen while you’re away from the terminal. Channels can be two-way: Claude reads an inbound event and replies back through the same channel. Telegram, Discord, and iMessage are included in the research preview. See §8 MCP.
Checkpoint
Section titled “Checkpoint”A restore point created at each prompt you send. Claude Code snapshots files before every edit, so you can revert code, conversation, or both to an earlier state. Press Esc twice or run /rewind to restore. Checkpoints are local to the session, separate from git, and do not track changes made through the Bash tool. See §9 Checkpointing.
.claude directory
Section titled “.claude directory”The directory where Claude Code reads project-scoped configuration: settings, hooks, skills, subagents, rules, and auto memory. A project has .claude/ at its root; your user-level defaults live at ~/.claude/. See §3 Configuration.
CLAUDE.md
Section titled “CLAUDE.md”A markdown file of persistent instructions you write for Claude, loaded at the start of every session as a user message after the system prompt. Put project conventions, architecture notes, and behavioral rules here. CLAUDE.md survives compaction and is re-read fresh from disk afterward. See §3.1 Memory Files.
Compaction
Section titled “Compaction”Automatic summarization of your conversation when the context window approaches its limit. Older tool outputs clear first, then the conversation is summarized. CLAUDE.md and auto memory survive compaction and reload from disk; instructions given only in conversation may be lost. Run /compact to trigger manually. See §2 Context Window.
Dispatch
Section titled “Dispatch”A phone-initiated task router that spawns a Claude Code session in the Desktop app when you send a coding task from the Claude mobile app. Your prompt routes to the right tool automatically. Available on Pro and Max plans. See §9 Desktop Features.
Effort level
Section titled “Effort level”A setting that controls how much of the adaptive-reasoning thinking budget Claude uses per turn. Higher effort means more thinking tokens and deeper reasoning; lower effort is faster and cheaper. Supported on Opus 4.6+ and Sonnet 4.6. Skills can override the session effort level in their frontmatter. See §2.5 Model Selection.
Extended thinking
Section titled “Extended thinking”Visible step-by-step reasoning the model performs before responding. You can cap thinking tokens with MAX_THINKING_TOKENS or adjust the effort level. Thinking appears in gray italic text in the terminal. See §2.5 Model Selection.
A user-defined handler that executes automatically at a specific point in Claude Code’s lifecycle, such as before a tool runs, after a file edit, or at session start. A hook configuration has three levels: the hook event (which lifecycle point), the matcher (which events fire it), and the hook handler (what runs). Handlers can be a shell command, HTTP endpoint, MCP tool, LLM prompt, or subagent. Hooks are deterministic: they fire at fixed lifecycle points, not at the model’s discretion. See §7 Hooks.
Managed settings
Section titled “Managed settings”A settings file enforced org-wide by IT or DevOps, placed at an OS-level path outside ~/.claude. Users cannot override or exclude managed settings. Use this for security policies, compliance requirements, or standardized tooling across a fleet. See §3.3 Settings.
MCP (Model Context Protocol)
Section titled “MCP (Model Context Protocol)”An open standard for connecting AI tools to external data sources and services. MCP servers give Claude new tools for Slack, Jira, databases, browsers, and hundreds of other integrations. Connect servers via /mcp or by adding them to .mcp.json. See §8 MCP.
MCP Tool Search
Section titled “MCP Tool Search”A context-saving mechanism that defers MCP tool schemas until needed. Only tool names load at startup; Claude fetches the full schema on demand when it decides to use a specific tool. This keeps idle MCP servers from consuming context. See §8 MCP.
Non-interactive mode
Section titled “Non-interactive mode”A mode that executes a single prompt and exits without a conversational session, invoked with -p or --print. Used for CI, scripts, and piping. The --bare flag extends this for clean-slate automation. Formerly called “headless mode.” See §9 Non-interactive Mode.
Output style
Section titled “Output style”A configuration that modifies Claude’s system prompt to change response behavior, tone, or format. Output styles turn off the software-engineering-specific parts of the default system prompt, unlike CLAUDE.md (which is delivered as a user message). Built-in styles include Default, Proactive, Explanatory, and Learning. See §3 Configuration.
Permission mode
Section titled “Permission mode”The baseline approval behavior for the session. Available modes are default, acceptEdits, plan, auto, dontAsk, and bypassPermissions. Cycle with Shift+Tab in the CLI or use the mode selector in VS Code, Desktop, and claude.ai. See §3.3 Settings.
Permission rule
Section titled “Permission rule”A settings entry that allows, asks about, or denies a tool invocation based on the tool name and argument pattern. Rules evaluate in deny-then-ask-then-allow order, so the first match wins. Permission rules are fine-grained controls layered on top of the broader permission mode. See §3.3 Settings.
Plan mode
Section titled “Plan mode”A permission mode where Claude researches and proposes changes without editing your source files. It can read, search, and run exploration commands, then presents a plan for approval before touching anything. Enter plan mode with /plan or by pressing Shift+Tab. See §3.3 Settings.
Plugin
Section titled “Plugin”A bundle of skills, hooks, subagents, and MCP servers packaged as a single installable unit. Plugin skills are namespaced as plugin-name:skill-name so multiple plugins coexist without conflicts. Distribute plugins across teams via a marketplace. See §5 Skills.
Project trust
Section titled “Project trust”A dialog that accepts a directory before Claude Code loads its configuration. Acceptance is saved per project directory, except your home directory (where trust is per-session only). Trust gates auto-installation of marketplace plugins and execution of project-defined hooks. See §3 Configuration.
Prompt injection
Section titled “Prompt injection”Hostile instructions embedded in a file, web page, or tool result that attempt to redirect Claude toward actions you never requested. Claude Code’s defenses include the permission system, command blocklists, and trust verification. Auto mode adds a server-side classifier that never sees tool results, so injected text cannot influence its approval decisions. See §10 Security.
Remote Control
Section titled “Remote Control”A way to continue a local Claude Code session from your phone or browser via claude.ai. Your code stays on your machine; only the UI is remote. Different from Claude Code on the web, which runs in a cloud sandbox. See §9.22 Remote Control.
Modular instruction files in .claude/rules/ that load alongside CLAUDE.md. A rule can be path-scoped with YAML paths: frontmatter so it only loads when Claude reads a matching file, keeping context lean until relevant. See §3.1 Memory Files.
Sandboxing
Section titled “Sandboxing”OS-level filesystem and network isolation for the Bash tool. Commands run inside a boundary you define upfront, so Claude can work freely within it without per-command approval prompts. Sandboxing is a separate layer from permission rules. See §10 Security.
Session
Section titled “Session”A conversation tied to your current directory, with its own independent context window. Sessions can be resumed with claude -c, forked with --fork-session, or run in parallel across terminals. Each session’s transcript is stored under ~/.claude/projects/. See §1 How Claude Code Works.
Settings layers
Section titled “Settings layers”The hierarchy Claude Code reads configuration from, in precedence order from highest to lowest: managed policy, command-line arguments, local settings at .claude/settings.local.json, project settings at .claude/settings.json, then user settings at ~/.claude/settings.json. Arrays merge across layers; scalars at a higher layer override lower ones. See §3.3 Settings.
A SKILL.md file containing instructions, knowledge, or a workflow that Claude adds to its toolkit. Invoke directly with /skill-name. Skills follow the Agent Skills open standard; Claude Code extends it with invocation control and subagent execution. Skills are the recommended successor to custom commands: .claude/commands/ files continue to work, but .claude/skills/ is the preferred location. See §5 Skills.
Subagent
Section titled “Subagent”A specialized AI assistant that runs in its own context window with a custom system prompt, specific tool access, and independent permissions. It works on a delegated task and returns a summary to the main conversation. Unlike agent teams (where each agent is a full independent session), subagents run within and report to the parent session. See §4 Agents.
Surface
Section titled “Surface”Any place you access Claude Code: the CLI, VS Code, JetBrains, Desktop, or claude.ai. All surfaces share the same engine, so your CLAUDE.md, settings, and skills work identically across them. Slack and the Chrome extension are integrations that connect to a surface rather than surfaces themselves.
Teleport
Section titled “Teleport”A command (/teleport) that pulls a cloud Claude Code session into your local terminal. Claude fetches the branch, loads the conversation history, and resumes from the web session’s last state. The reverse direction is --remote, which sends a local task to run on the web. See §9.16 Session Teleportation.
An action Claude can take: read a file, edit code, run a shell command, search the web, spawn a subagent. Tools are what make Claude Code agentic. Without them, Claude can only respond with text. Each tool use returns a result that informs Claude’s next decision in the agentic loop. See §1 How Claude Code Works.
One complete response from Claude within a session. A turn begins when you send a message and ends when Claude finishes responding, with any number of tool calls in between. Stop hooks fire at the end of each turn. See §7 Hooks.
Verification loop
Section titled “Verification loop”A session pattern where you give Claude a check it can run (a test suite, a build, or a screenshot comparison) and Claude iterates until the check passes instead of stopping after one attempt. A verification loop is the prerequisite for /goal, unattended runs, and dynamic workflows: without one, the only signal that the agent is finished is the agent itself. See §9 Advanced Features.
Worktree isolation
Section titled “Worktree isolation”An isolation mode that runs Claude in a separate git worktree under .claude/worktrees/, enabled with the -w flag or isolation: worktree in subagent config. Changes stay on a separate branch and directory, so parallel agents cannot overwrite each other’s files. See §4 Agents.
Deprecated and renamed terms
Section titled “Deprecated and renamed terms”These terms appear in older documentation and community content. Use the current name when searching this guide.
| Old term | Now called | Notes |
|---|---|---|
| Headless mode | Non-interactive mode | Same -p flag, same behavior |
| Custom commands | Skills | .claude/commands/ files still work |
| Slash commands | Commands | ”Slash” dropped from product copy |