Claude Code FAQ
Claude Code FAQ hub for install, updates, pricing, limits, CLAUDE.md memory, Plan Mode, MCP, ultrathink, ultracode, custom commands, headless mode, and API docs.
This FAQ hub is the short-answer layer of ClaudeCode101. Use it when you have a specific Claude Code question and need a fast answer before opening the full tutorial, mechanics guide, MCP page, pricing note, or troubleshooting entry.
Last checked on July 7, 2026. Claude Code installation methods, package manager behavior, plan access, model availability, dynamic workflows, and usage limits change often. Treat this page as a routing map, then verify account-specific pricing, usage, and model access in the official Claude surfaces before changing a production workflow.
Quick Answer
Start with the question that matches your intent:
- Installing and updating Claude Code: use the install FAQ first, then check whether your update should use
claude update, npm, Homebrew, WinGet, apt, dnf, or apk. - Windows setup: decide between native Windows and WSL before changing PATH or shell configuration.
- Pricing and limits: check pricing, usage limits,
/model,/usage,/clear, and/compacttogether. - Project memory: read the CLAUDE.md FAQ and the official memory docs FAQ before creating a large project instruction file.
- Plan Mode: use it before risky multi-file edits, migrations, pricing pages, redirects, auth, or payments.
- MCP servers: connect one server at a time, with narrow permissions.
- Automation: use custom commands and skills for repeatable prompts, the arguments FAQ for command inputs, headless mode for scripts, and Agent SDK/API docs for programmatic boundaries.
- Deep reasoning: compare Ultrathink, Ultracode, and Plan Mode before spending extra latency or token budget.
The FAQ page should answer the first question. The linked page should help you act.
Start Here
| Question | Short answer | Open next |
|---|---|---|
| How do I install Claude Code? | Install with the current official path, verify the claude command, then authenticate. | Installation tutorial |
| Claude update vs npm install: which should I use? | Update with the same channel that installed your active claude binary. | Latest version and changelog |
| How do I install Claude Code on Windows? | Use native Windows for Windows projects; use WSL 2 when your toolchain is Linux-oriented. | Windows PATH troubleshooting |
| How much is Claude Code? | Subscription users should compare Free, Pro, Max, Team, and Enterprise; API users should think in token cost. | Pricing and limits hub |
| What are Claude Code usage limits? | Limits depend on how you sign in, model choice, context size, and plan/API billing surface. | Rate limit troubleshooting |
| What is CLAUDE.md? | It is durable project memory for commands, conventions, architecture boundaries, and review rules. | CLAUDE.md mechanics |
| Where are Claude Code memory official docs? | Use official memory docs for behavior, then use ClaudeCode101 for setup choices. | CLAUDE.md mechanics |
| What is Plan Mode? | It tells Claude Code to inspect and propose a plan before editing source files. | Plan Mode mechanics |
| What are Claude Code MCP servers? | MCP servers add external tools such as GitHub, browser automation, docs, databases, and internal systems. | MCP directory |
| What are Claude Code custom commands? | Custom commands and slash commands turn repeatable prompt patterns into named workflows. | Slash commands mechanics |
| How do custom command arguments work? | Text after the slash command name becomes arguments; use skills for structured inputs. | Custom commands tutorial |
| What is Claude Code headless mode? | Headless mode runs Claude Code non-interactively for scripts, JSON output, checks, and CI-style flows. | API documentation FAQ |
| Where is the Claude Code API documentation? | Use CLI docs for terminal workflows, Agent SDK docs for agent automation, and Claude API docs for model calls. | Headless mode FAQ |
| Ultrathink vs Ultracode: what is the difference? | ultrathink helps one hard reasoning step; Ultracode is for larger scoped repository work. | Ultrathink guide |
Installation And First Run Questions
Installation questions usually split into two decisions: operating system and authentication path.
| If you are asking... | Read this first | What to verify |
|---|---|---|
| "How do I install Claude Code today?" | How to install Claude Code | claude --version, shell PATH, and sign-in method. |
| "How do I update Claude Code?" | Claude update vs npm install | Active binary path, package manager, and current dist-tags. |
| "Why does Windows behave differently?" | Windows install guide | Native terminal vs WSL, package manager, and PATH. |
"Why does claude not run?" | Command not found fix | The binary location and the shell that loads your PATH. |
| "Why can I not log in?" | Login troubleshooting | Browser auth, API key, organization seat, and local credential state. |
Do not troubleshoot pricing, MCP, or Plan Mode until the CLI is installed, visible in PATH, and authenticated.
Pricing, Usage, And Cost Questions
Claude Code pricing questions can be confusing because subscription usage and API billing are different surfaces.
| Question | Practical answer |
|---|---|
| "Is Claude Code free?" | Claude plan availability can change; check the official pricing page and your account. Some workflows may require paid access or API billing. |
| "Should I use Pro, Max, Team, or Enterprise?" | Choose based on coding frequency, team needs, admin/security requirements, and whether you need organization-level controls. |
| "Why did I hit a limit?" | Check sign-in method, active model, context size, long sessions, and whether you are on subscription usage or API-style metering. |
| "How do I stretch usage?" | Use /clear between unrelated tasks, /compact for long sessions, /model for model choice, and smaller file-focused prompts. |
| "How do I track cost?" | Subscription users should inspect plan usage surfaces; API users should use /usage, /cost, and the Console or cloud-provider billing page. |
| "Why did Opus 4.8 feel slow or expensive?" | Check effort, ultracode, dynamic workflows, context size, and Fast mode before blaming the model. |
For production budgeting, start from How much is Claude Code?, then read Claude Code usage limits.
Workflow And Mechanics Questions
Mechanics pages explain why a workflow works and what to avoid.
| Workflow need | Start with | Then read |
|---|---|---|
| Keep project rules stable | Claude Code memory official docs | CLAUDE.md mechanics |
| Make Claude plan before edits | What is Plan Mode? | Plan Mode mechanics |
| Reduce context bloat and quota waste | Usage limits | Context management |
| Reuse a repeated prompt | Custom command arguments | Custom commands tutorial |
| Add automatic guardrails | Custom commands | Hooks |
| Delegate noisy side work | API documentation | Subagents |
| Orchestrate a large audit or migration | Latest version and changelog | Dynamic workflows |
| Choose deep reasoning or workflow effort | Ultrathink vs Ultracode | Ultracode |
If the question is "what should I configure first," the usual order is CLAUDE.md, permissions, one reusable command, Plan Mode habits, then hooks or subagents only after the pattern repeats. Dynamic workflows come later, when the task is broad enough to justify extra usage and orchestration.
MCP, API, And Automation Questions
Use these questions when Claude Code needs to leave the basic terminal workflow.
| Need | Best entry point | Watch out for |
|---|---|---|
| Connect GitHub, browser, docs, or databases | MCP servers FAQ | Start with one server and narrow tool permissions. |
| Choose an MCP server | MCP directory | Prefer trusted sources and explicit scopes. |
| Run scripted prompts | Headless mode FAQ | Make permissions, credentials, and output format explicit. |
| Build custom automation | API documentation FAQ | Do not mix CLI, Agent SDK, and Claude API docs without a clear boundary. |
| Keep MCP safe | MCP security checklist | External tools can read or write real systems. |
MCP and automation are where mistakes become more expensive. Treat tool access as a production decision, not a convenience toggle.
How To Use This FAQ
Use a FAQ page for the first answer, not as the final manual. A strong question page should do five things:
- Answer the question in the first few lines.
- State what changes often and must be verified.
- Give the practical next action.
- Link to the deeper tutorial or mechanics page.
- Link to official sources where account-specific details live.
Before making a production decision, verify:
- Current plan availability and billing rules.
- Which account, organization, or API key Claude Code is using.
- Which model is active in
/model. - Whether the task should run interactively, in Plan Mode, headless mode, or the Agent SDK.
- Whether tool permissions and MCP servers are narrow enough for the repository.
- Whether the page you read has a recent "Last checked" date.
Related Hubs
- Claude Code tutorial center
- Claude Code mechanics
- MCP directory
- Troubleshooting
- Pricing and limits
- Resources and tools