FAQ

Frequently asked questions about Claude Code

Frequently Asked Questions (FAQ)

This document answers common questions about using Claude Code.

Installation & Setup

Q: How do I install Claude Code?

A: Claude Code is installed globally via npm:

npm install -g @anthropic-ai/claude-code

After installation, run claude in any project directory to start.

Q: The claude command is not found after installation?

A: This usually means npm's global directory is not in your PATH. Solution:

# Check npm global directory
npm config get prefix

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

Q: How do I configure QCode.cc proxy?

A: Set these environment variables:

export ANTHROPIC_BASE_URL="https://asia.qcode.cc/api"
export ANTHROPIC_AUTH_TOKEN="cr_your_api_key"

We recommend adding these to your shell config file (e.g., ~/.zshrc).


API Key & Authentication

Q: Where do I get an API Key?

A: Log in to QCode.cc Dashboard and create or view your API Key in the API Key management page.

Q: What prefix does the API Key have?

A: QCode.cc API Keys start with cr_, e.g., cr_xxxxxxxxxxxxxx.

Q: Why is my API Key not working?

A: Possible reasons: 1. Copy error: Check for extra spaces or missing characters 2. Subscription expired: Check your account subscription status 3. Environment variable not loaded: Reopen terminal or run source ~/.zshrc 4. Wrong API endpoint: Confirm ANTHROPIC_BASE_URL is set correctly


Usage & Costs

Q: How do I check current session usage?

A: There are two ways to check usage:

Option 1: Dashboard (Recommended)

Log in to QCode.cc Dashboard and check the "Usage Statistics" section:

  • Model Call Counts: Usage frequency for each model
  • Cost Breakdown: Accurate cost consumption details
  • Quota Usage: Current plan quota consumption progress

Dashboard data updates in near real-time, with more accurate billing. We use the industry-standard LiteLLM pricing rules and sync updates regularly to ensure billing accuracy.

Option 2: Quick Check in CLI

Use the /cost command for a quick overview of current session:

/cost

Example output:

Total cost:            $0.55
Total duration (API):  6m 19.7s
Total code changes:    10 lines added, 5 lines removed

Note: The cost shown by /cost is for reference only, useful for quickly checking model distribution and approximate usage. Since Anthropic hasn't published detailed pricing rules for each model, the CLI's built-in billing may differ from actual costs. We recommend using Dashboard statistics as the authoritative source.

Q: How can I reduce token consumption?

A: Several effective methods:

  1. Compress context: Use /compact command
  2. Clear history: Use /clear to reset context
  3. Be specific: Avoid vague questions that cause unnecessary scanning
  4. Break down tasks: Split large tasks into smaller ones

Q: When does context auto-compact?

A: Claude Code auto-compacts by default when context reaches 95% capacity. You can also: - Manually run /compact to compact proactively - Add custom instructions after /compact to control what to focus on


Permission System

Q: What permission modes are available?

A: Four permission modes:

Mode Description
default Prompts on first use of each tool
acceptEdits Auto-approves file edits
plan Analyze only, no modifications (Plan Mode)
bypassPermissions Skips all prompts (requires safe environment)

Q: How do I set a default permission mode?

A: Configure in .claude/settings.json:

{
  "permissions": {
    "defaultMode": "plan"
  }
}

Q: How do I prevent Claude from reading sensitive files?

A: Configure deny rules in .claude/settings.json:

{
  "permissions": {
    "deny": [
      "Read(./.env)",
      "Read(./.env.*)",
      "Read(./secrets/**)"
    ]
  }
}

Q: How do I pre-approve common commands?

A: Configure allow rules in .claude/settings.json:

{
  "permissions": {
    "allow": [
      "Bash(npm run lint)",
      "Bash(npm run test:*)",
      "Bash(git status)"
    ]
  }
}

CLAUDE.md Configuration

Q: What is CLAUDE.md?

A: CLAUDE.md is a project-level config file that provides project context to Claude, such as: - Project architecture - Coding conventions - Common commands - Special conventions

Q: Where should I put CLAUDE.md?

A: Place it in the project root. Claude Code reads it automatically on startup.

Q: How do I create CLAUDE.md?

A: Use the /init command to auto-generate:

/init

Or create manually with this basic structure:

# Project Name

## Overview
Brief description of the project...

## Tech Stack
- Frontend: React
- Backend: FastAPI

## Common Commands
- Start dev server: `npm run dev`
- Run tests: `npm test`

## Coding Standards
- Use TypeScript
- Prefer functions over classes

Context Management

Q: How do I clear session history?

A: Use the /clear command:

/clear

This clears all conversation history and starts a fresh session.

Q: How do I compact without clearing history?

A: Use the /compact command:

/compact

You can add instructions to control what to focus on:

/compact Preserve all code examples

Q: How do I check current context status?

A: Use the /context command:

/context

Q: How do I use @ to reference files?

A: Use @ to directly reference files into context:

@src/main.ts Explain this file
@package.json Check dependency versions

MCP Servers

Q: What is MCP?

A: MCP (Model Context Protocol) is a protocol that allows Claude Code to connect to external tools and data sources, extending its capabilities.

Q: How do I configure MCP servers?

A: Configure in ~/.claude/settings.json:

{
  "mcpServers": {
    "context7": {
      "command": "npx",
      "args": ["-y", "@context7/mcp-server"]
    }
  }
}

Q: Why is my MCP server connection failing?

A: Check the following: 1. Is the server command correct? 2. Are required dependencies installed? 3. Are environment variables properly set? 4. Use /mcp command to check status


Common Operations

Q: How do I enter Plan Mode?

A: Press Shift+Tab twice, or set it as default mode in settings. In Plan Mode, Claude analyzes but doesn't execute.

Q: How do I switch models?

A: Use the /model command:

/model opus
/model sonnet

Q: How do I see available commands?

A: Use /help command to see all built-in commands.

Q: How do I interrupt a running operation?

A: Press Escape to interrupt the current operation.

Q: How do I make Claude remember certain information?

A: Use the /memory command to add to session memory:

/memory This project uses pnpm instead of npm

Git Operations

Q: How do I get Claude to commit code?

A: Use the /commit command:

/commit

Claude will analyze changes and generate a commit message.

Q: How do I get Claude to review code?

A: Use the /review command:

/review

You can specify a scope:

/review src/auth/

Troubleshooting

Q: Claude is responding slowly?

A: Try these methods: 1. Check network connection 2. Use /compact to compress context 3. Avoid peak hours 4. Check if large files are in context

Q: Getting "Rate limit exceeded"?

A: This is a 429 error indicating too many requests: 1. Wait 30-60 seconds before retrying 2. Reduce request frequency 3. See Error Code Reference for details

Q: Claude can't read a file?

A: Check: 1. Does the file exist? 2. Are file permissions correct? 3. Is the file in a deny rule? 4. Is the path correct?

Q: How do I report a bug?

A: 1. Visit GitHub Issues 2. Provide detailed reproduction steps 3. Include relevant error messages


More Help

If your question isn't answered here:

  1. Official docs: Claude Code Documentation
  2. QCode.cc support: Live chat at bottom right of website
  3. Community: GitHub Discussions