Skip to main content

โ“ Frequently Asked Questions

Installation and Getting Startedโ€‹

Q. I installed Claude Code, but the claude command is not found.โ€‹

Your npm global install path is not on your PATH.

# Check the global prefix
npm config get prefix

# Add to ~/.bashrc or ~/.zshrc
export PATH="$PATH:$(npm config get prefix)/bin"

# Apply
source ~/.bashrc

Q. I get an error saying my Node.js version is incompatible.โ€‹

Claude Code requires Node.js 18 or later.

node --version  # Check your current version

# Upgrade with nvm
nvm install 20
nvm use 20

Q. I get a permission error on macOS.โ€‹

To install without sudo, fix the ownership of the npm global directory.

sudo chown -R $(whoami) $(npm config get prefix)/{lib/node_modules,bin,share}
npm install -g @anthropic-ai/claude-code

API Keysโ€‹

Q. Where do I get an API key?โ€‹

Go to console.anthropic.com โ†’ API Keys. You need to sign up for an account and register a payment method.

Q. I set my API key but keep getting authentication errors (401).โ€‹

  1. Check for leading/trailing whitespace in the key
  2. Check for typos in the ANTHROPIC_API_KEY environment variable name
  3. Verify in the dashboard that the key is in "active" status
  4. If it still fails, issue a new key and try again

Q. Can I share an API key with teammates?โ€‹

Not recommended for security reasons. Have each person issue their own key, or split organization API keys by purpose. If a key leaks, deactivate it immediately and issue a new one.

Pricing and Costsโ€‹

Q. How much does Claude Code cost?โ€‹

Claude Code can be used in two ways:

  1. Subscription plans (Pro/Max/Team): usage included in a monthly flat fee
  2. API key: pay-as-you-go billing based on Anthropic API usage

Check current pricing on the Anthropic pricing page.

Q. I don't know how much I'll spend per month. Can I set a cap?โ€‹

You can set a monthly spend limit (Usage Limit) in the Anthropic Console. Go to console.anthropic.com โ†’ Settings โ†’ Limits.

Q. Can I check how much I've spent so far during a session?โ€‹

Type the /cost command during a Claude Code conversation to see the cumulative token usage and estimated cost for the current session.

Errors During Useโ€‹

Q. I get a "Context window exceeded" error.โ€‹

The context (conversation content) has exceeded the model's maximum length.

  • Summarize the conversation with the /compact command, or
  • Start fresh with /clear
  • Storing key information in CLAUDE.md lets you keep context across new sessions

Q. Claude keeps repeating the same mistakes.โ€‹

Add explicit "do not" rules to CLAUDE.md.

## Cautions
- Never modify `package-lock.json` manually
- Do not delete test files
- Do not include `console.log` debug code in commits

Q. I modified a file and want to revert it.โ€‹

Files modified by Claude Code are tracked by git.

git diff          # Review the changes
git restore .     # Revert everything
git restore <file>  # Revert a specific file only

Q. Claude stops responding mid-answer.โ€‹

  • This can happen when the network is unstable
  • If the response is very long, break the task into smaller requests
  • Run the /doctor command to diagnose your installation and connection status

IDE Integrationโ€‹

Q. I can't find the Claude Code extension in VS Code.โ€‹

Search for "Claude Code" in the VS Code Extensions marketplace, or run claude in the terminal and choose the IDE integration option.

Q. Can I use it with JetBrains IDEs (IntelliJ, PyCharm, etc.)?โ€‹

Install the Claude Code plugin from the JetBrains plugin marketplace. It connects with the same API key.

Agent SDKโ€‹

Q. How do I use the Agent SDK?โ€‹

The Agent SDK is a tool for running Claude Code programmatically. There are three ways to use it:

  1. CLI: claude -p "prompt" (simplest)
  2. Python SDK: pip install claude-code, then call it from Python
  3. TypeScript SDK: install the npm package, then call it from TypeScript

For details, see the official Agent SDK documentation.

Q. Does Claude Code need to be installed to run the Agent SDK?โ€‹

Yes. The Agent SDK uses Claude Code's agentic loop. It is not a direct API call โ€” it controls Claude Code programmatically.

Q. Can I run multiple agents concurrently with a single API key?โ€‹

Yes, you can. However, you may hit rate limits depending on the number of concurrent requests, so apply queuing if you run many agents.

Miscellaneousโ€‹

According to Anthropic's terms of service, ownership of the output you generate belongs to you. That said, always consult a legal professional on licensing matters.

Q. Can I use it offline?โ€‹

Claude Code requires an internet connection to the Anthropic API. It cannot be used in fully offline environments.

Q. If I ask in Korean, does it answer in Korean?โ€‹

Yes, Claude fully supports Korean. If you ask in Korean it answers in Korean, and adding Primary Language: Korean to CLAUDE.md makes responses more consistently Korean.

Q. Did the /review command disappear?โ€‹

No, it's available again as a built-in command. As of v2.1.202, /review is back as a fast single-pass review, and multi-agent review has been split into a separate /code-review command:

/review <pr#>               # Fast single-pass review
/code-review <level> <pr#>  # Multi-agent review with a specified effort level

As of v2.1.215, Claude no longer runs /code-review (or /verify) on its own, so invoke them yourself when needed.

Q. What is Fast Mode?โ€‹

A high-speed mode toggled with /fast. It runs the same model faster and is billed separately. It's well suited for live debugging or rapid iteration. You can disable it with CLAUDE_CODE_DISABLE_FAST_MODE=1.

Q. How do I diagnose my installation?โ€‹

Run the /doctor command to automatically diagnose your Claude Code installation, configuration, and connection status.

Last updated on