Claude Code 2025: Complete Guide (Setup, Commands, Pricing)

What Is Claude Code?

Claude Code is Anthropic's official agentic coding tool. Unlike Cursor (a GUI IDE) or GitHub Copilot (an autocomplete extension), Claude Code is designed as a terminal-first agent that can:

• Read your entire codebase (up to 200k tokens of context)
• Edit multiple files in a single session
• Run shell commands, tests, and git operations
• Use MCP (Model Context Protocol) servers for external tool integrations
• Work inside VS Code and JetBrains IDEs via the Claude Code extension

It's built by Anthropic and runs exclusively on Claude models (claude-sonnet-4-6, claude-opus-4, etc.). Claude Code is the reference implementation of how to build an agentic coding workflow on top of Claude.

Installation

Claude Code is distributed as an npm package. You need Node.js 18+.

# Install globally
npm install -g @anthropic-ai/claude-code

# Verify installation
claude --version

# Start a session
claude

On first launch, Claude Code opens a browser window to authenticate with your Anthropic account. Alternatively, set the environment variable:

export ANTHROPIC_API_KEY=sk-ant-...

Key Slash Commands

Claude Code has a set of built-in slash commands for session management and workflow control:

You can also define custom slash commands in your CLAUDE.md or project configuration for common workflows (e.g., /deploy, /test-all).

Keyboard Shortcuts

Setting Up CLAUDE.md

CLAUDE.md is the most important configuration file for Claude Code. It acts as a persistent system prompt injected at the start of every session in that directory. Run /init to auto-generate one, or write it manually:

# CLAUDE.md example
You are working on a TypeScript + React SPA.
- Use functional components with typed props (no 'any')
- Test with Vitest + Testing Library
- Prefer named exports; Tailwind for styling
- Never commit secrets or .env files

Claude Code reads CLAUDE.md from multiple locations in order:

1. Your home directory ~/.claude/CLAUDE.md (global rules)
2. Your project root ./CLAUDE.md (project rules)
3. Subdirectory CLAUDE.md files (module-specific rules)

MCP Integrations

Claude Code supports MCP (Model Context Protocol) servers, letting you connect Claude to external tools: databases, GitHub, Jira, file systems, browser automation, and more. MCP servers are configured in ~/.claude/settings.json:

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_TOKEN": "ghp_..." }
    }
  }
}

Once configured, Claude can call GitHub APIs, query databases, or control a browser — all within a single Claude Code session. Explore available MCP servers at prismix.dev/mcp.

Pricing (2025)

For occasional use, Pro ($20/mo) is sufficient. Heavy users who run multi-hour coding sessions should evaluate Max — the per-session cost of API tokens adds up quickly at scale.

Claude Code vs Cursor vs GitHub Copilot

Claude Code is the best choice for developers who want a terminal-native, deeply agentic workflow with full shell access and MCP extensibility. For more, see our Cursor guide and GitHub Copilot guide. If you're choosing between the two, read Claude Code vs Cursor.

Power User Tips

• Use /compact aggressively: Long sessions fill the 200k context window. Run /compact after each major task to compress history and preserve working context for the rest of the session.
• Prefix tasks with context: Start messages with "In [filename], ..." or "Looking at [function]..." — Claude responds better to scoped requests than vague ones.
• Use headless mode for scripting: Run claude -p "your prompt" --output-format json to integrate Claude Code into CI pipelines or shell scripts.
• Chain MCP + shell: Claude Code can call an MCP server (e.g., GitHub) and then run a local shell command in the same turn — making it more powerful than either alone.
• Monitor Anthropic uptime: If Claude Code stops responding, check Anthropic API status at Prismix before debugging locally.