Claude Code glossary: 60 terms in plain English
Claude Code glossary: 60 terms in plain English
This glossary defines the Claude Code terms you will meet in docs, proposals and team chats, in plain English. Each entry says what the thing is and, where it matters, why a business website team should care. The definitions follow Anthropic's own documentation as of October 2026.
Terms are grouped by topic rather than alphabet, so related ideas sit together: the basics, instructions and memory, permissions and safety, extensions, context, automation, surfaces and plans. Anthropic renames things often. "Headless mode" is now non-interactive mode, "slash commands" are now just commands, and custom commands have been folded into skills. Where a term has an old name, both are given.
Key Takeaways
- Claude is the AI model. Claude Code is the agent software around it, which Anthropic calls the "harness."
- CLAUDE.md is the file of standing instructions you write; auto memory is the notes Claude writes for itself.
- Permission modes set what Claude can do without asking. As of October 2026, current versions start interactive terminal and VS Code sessions in auto mode, where a classifier model reviews actions instead of you.
- Skills, hooks, subagents and MCP servers are the four main ways to extend Claude Code. Hooks are the only one that runs every time without the model deciding.
- Checkpoints undo local file edits, but they can't undo deployments, database changes or other actions on remote systems.
The basics
| Term | Plain-English meaning |
|---|---|
| Claude Code | Anthropic's agentic coding tool. It "reads your codebase, edits files, runs commands, and integrates with your development tools." |
| Claude (the model) | The AI model that does the reasoning. Current models include Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1, as of October 2026. |
| Agentic coding | Work where the AI reads files, runs commands and makes changes itself, instead of only replying with text you apply by hand. |
| Agentic harness | The software that turns a model into an agent. Anthropic: "Claude Code is the harness; Claude is the model inside it." |
| Agentic loop | The cycle Claude repeats on every task: gather context, take action, verify results. You can interrupt it at any point. |
| Tool | One action Claude can take: read a file, edit code, run a shell command, search the web, start a subagent. |
| Session | One conversation tied to a project folder, with its own memory of what was said. You can resume an old one or start fresh. |
| Token | A small chunk of text the model reads or writes. API prices are quoted per million tokens. |
| Effort level | How hard the model thinks. "Higher effort means more thinking tokens and deeper reasoning; lower effort is faster and cheaper." |
| Verification loop | Giving Claude a check it can run, such as a build, a test or a screenshot comparison, so it keeps working until the check passes instead of stopping at "looks done." |
For a fuller introduction, see what Claude Code is, and for how it differs from the chat app and developer API, see Claude.ai vs Claude Code vs the Claude API.
Instructions and memory
| Term | Plain-English meaning |
|---|---|
| CLAUDE.md | A Markdown file of standing instructions you write, loaded at the start of every session: project conventions, build commands, "always do X" rules. The /init command writes a starter version. It can sit in the project (./CLAUDE.md), in your user folder, or be set by your organization. |
| AGENTS.md | A shared instructions file some other coding agents use. If a project has one and no CLAUDE.md, Claude reads it instead. |
| Auto memory | Notes Claude writes for itself from your corrections and preferences. The first 200 lines or 25 KB of its index load each session. |
| Rules | Extra instruction files in .claude/rules/ that load alongside CLAUDE.md. A rule can apply only to certain files, such as blog templates. |
| .claude directory | The project folder where Claude Code keeps its settings, hooks, skills, subagents and rules. Your personal defaults live in ~/.claude/. |
| Settings files | JSON files that configure Claude Code. Highest priority wins: organization policy, then command-line settings, then .claude/settings.local.json, then .claude/settings.json, then your user settings. |
| Output style | A set of instructions that changes Claude's tone and response format for a whole session. A custom one can make Claude act as something other than a software engineer, such as a writing assistant. |
Writing a good CLAUDE.md for a website is its own skill. Our guide on how to write a CLAUDE.md for a website covers what goes in and what stays out, and our Claude Code templates page has copy-ready files.
Permissions and safety
| Term | Plain-English meaning |
|---|---|
| Permission mode | The baseline for what Claude can do without asking. Switch with Shift+Tab in the terminal or the mode selector in the apps. |
| Manual mode | Claude asks before file edits and shell commands. Its config value is default. Best for sensitive work and for learning. |
| Accept edits | Claude edits files and runs common file commands without asking, but still asks before other commands. |
| Plan mode | Claude researches and proposes a plan without editing your source files, and waits for your approval. Enter with /plan. |
| Auto mode | A separate classifier model reviews Claude's actions instead of you and blocks risky ones. The built-in starting mode for interactive terminal and VS Code sessions from version 2.1.283. |
| Don't ask mode | Only pre-approved tools run; anything that would prompt is denied. Anthropic suggests it for "Locked-down CI and scripts." |
| Bypass permissions | Skips permission checks entirely. Anthropic reserves it for "Isolated containers and VMs only" and warns it "offers no protection against prompt injection or unintended actions." Never use it on a real site. |
| Permission rule | A setting that allows, asks about or denies a specific tool or command, such as blocking reads of an environment file. Deny rules apply in every mode. |
| Sandboxing | An operating-system boundary around the shell commands Claude runs, limiting which files and network addresses they can reach. Available on macOS, Linux and WSL2, not native Windows. |
| Checkpoint | A restore point created at each prompt you send, with files snapshotted before every edit. Press Esc twice or run /rewind to go back. Separate from git, and it doesn't cover changes made through shell commands. |
| Prompt injection | Hostile instructions hidden in a file, web page or tool result that try to steer Claude toward actions you never asked for. |
| Managed settings | Settings an organization's IT team enforces for everyone. User and project settings can't override them. |
How these fit together on a live business site is covered in our guide to whether Claude Code is safe for production websites.
Extending Claude Code
| Term | Plain-English meaning |
|---|---|
| Command | A reusable instruction you run by typing /name, such as /clear or /model. Formerly "slash commands." |
| Skill | A SKILL.md file holding instructions or a workflow, such as "run our SEO checks." Claude loads it when relevant, or you call it with /skill-name. The recommended successor to custom commands. |
| Hook | A handler that runs automatically at a fixed point, such as before a tool runs or after a file edit. Hooks are deterministic: they always fire, unlike CLAUDE.md instructions, which are advisory. |
| Subagent | A helper with its own context window, instructions and tool access. It does a delegated task and returns a summary. Built-ins include Explore, Plan and general-purpose. |
| Agent teams | Several Claude Code sessions coordinated by a lead with a shared task list. Experimental and off by default, and they use far more tokens. |
| MCP (Model Context Protocol) | "An open standard for connecting AI tools to external data sources and services." It is how Claude Code talks to tools like Notion, Slack or a database. |
| MCP server | A program that gives Claude new tools over MCP. Anthropic "does not security-audit or manage any MCP server," so vet each one. |
| Connector | An MCP server added to your claude.ai account, which then appears in Claude Code when you sign in with that account. |
| Plugin | A bundle of skills, hooks, subagents and MCP servers installed as one unit. |
For the MCP servers that matter to a web team, see MCP servers explained for website and marketing teams.
Context and long sessions
| Term | Plain-English meaning |
|---|---|
| Context window | The session's working memory: the conversation, files read, command output, CLAUDE.md and instructions. Anthropic notes performance degrades as it fills. Run /context to see what is using space. |
| Compaction | Automatic summarizing of the conversation when the context window is nearly full. Instructions given only in chat may be lost, which is why lasting rules belong in CLAUDE.md. Run /compact to do it yourself. |
| /clear | Starts a new session. The old one stays stored and can be reopened with /resume. Use it between unrelated jobs. |
Automation and parallel work
| Term | Plain-English meaning |
|---|---|
| Non-interactive mode | Runs a single prompt and exits, using -p. Used in scripts and CI. Formerly "headless mode." |
| Worktree | A separate copy of the project on its own branch, under .claude/worktrees/, started with -w. Parallel sessions each edit their own copy instead of overwriting each other. |
| Cloud session | A session that runs on cloud machines and keeps going after you close your laptop. Started from claude.ai/code, the mobile app, the desktop app or claude --cloud. |
| Remote Control | Steering a session running on your own computer from your phone or browser. Your files stay on your machine. |
| Routines | Scheduled or triggered cloud runs, created with /schedule. In research preview as of October 2026, with a minimum interval of one hour. |
| GitHub Actions integration | Runs Claude Code in your GitHub pipeline. Mention @claude in a pull request or issue to have it review or make changes. |
| Code Review | Anthropic's automatic pull request review, in research preview on Team and Enterprise plans as of October 2026. |
Using these for site checks before deploy is covered in Claude Code in GitHub Actions for website QA.
Surfaces and related products
| Term | Plain-English meaning |
|---|---|
| Surface | Any place you use Claude Code: the terminal, VS Code, JetBrains, the desktop app or claude.ai. All share the same engine. |
| CLI | Command-line interface: Claude Code in the terminal. The most complete surface, and the only one for scripting. |
| Desktop app (Code tab) | Claude Code inside the Claude desktop app, with visual diff review and no terminal. Requires a paid subscription. |
| IDE extension | Claude Code inside a code editor: VS Code (which also installs in Cursor) or JetBrains IDEs. |
| Claude Code on the web | The browser surface at claude.ai/code for cloud sessions. It clones code from GitHub. |
| Claude in Chrome | A browser extension that lets Claude use your logged-in browser to test pages or fill forms. |
| Claude Cowork | Claude's mode for delegated work, in the desktop, web and mobile apps on paid plans, using "the same agentic architecture that powers Claude Code, with no terminal required." |
| Agent SDK | A Python and TypeScript library with the same tools and agent loop as Claude Code, for building your own agents. |
Plans and billing
| Term | Plain-English meaning |
|---|---|
| Usage limits | On Pro and Max, a five-hour session limit and a weekly limit, shared between Claude chat and Claude Code. On Team, each seat has its own allowance on the same five-hour and weekly windows. |
| Usage credits | Pay-as-you-go usage after you reach your plan's limits, billed at standard API rates. You can set a monthly spend limit or choose unlimited. |
| Claude Console | Anthropic's developer account for API keys and per-token billing. Claude Code can run on it instead of a subscription. |
| /usage | Shows session cost, plan usage limits and activity. /cost and /stats are aliases. |
Conclusion
You don't need all sixty terms on day one. For a website team, the ones that matter first are CLAUDE.md, permission modes (especially plan mode and Manual mode), checkpoints and their limits, and the context window. The rest become useful as your setup grows. Anthropic's own Claude Code glossary links each term to its full docs page. If you want help setting up CLAUDE.md, permissions, hooks and MCP for your team, our Claude Code consulting covers it. More guides are in our Claude Code Guide.
Frequently asked questions
What is the difference between a skill and a hook?
A skill is a set of instructions Claude chooses to use when relevant, or that you call by name. A hook runs automatically at a fixed point every time, without the model deciding. Use a hook for anything that must always happen.
What is the difference between CLAUDE.md and auto memory?
You write CLAUDE.md. Claude writes auto memory, based on your corrections and preferences. Both load at the start of a session.
Is plan mode the same as Manual mode?
No. Plan mode lets Claude explore and propose a plan without editing your source files. Manual mode lets Claude edit, but only after you approve each edit and command.
What happened to headless mode and slash commands?
They were renamed. Headless mode is now non-interactive mode, run with -p. Slash commands are now called commands, and custom commands have been folded into skills, though old command files still work.
Does a checkpoint undo a deployment?
No. Checkpoints restore local file edits. Anthropic says actions on remote systems, such as databases, APIs and deployments, can't be checkpointed. Use git and your host's rollback for those.
Put this into action with eSEOspace
We help businesses grow with seo, geo & aeo that actually performs. Explore the services behind this guide:
Get a FREE Audit
We'll perform a comprehensive SEO, AEO, GEO & CRO audit of your website — completely free — and show you exactly how to outrank your competitors.
Don't have a site yet? Get in touch →
Get a FREE GEO/AEO/SEO Audit
We'll analyze your site's SEO, GEO, AEO & CRO — completely free — and show you exactly how to get found across Google and AI answers.
Don't have a site yet? Get in touch →
Great — your audit is on the way!
We'll send your free SEO/GEO/AEO/CRO audit within the next few hours. Where should we send it?
You're all set! ✓
Your free audit is being prepared — check your inbox in the next few hours. Talk soon!
