What it is
Claude Code is the official Anthropic CLI agent for coding. It runs inside your terminal — any shell, any project — and gives Claude full access to read files, run commands, edit code, and orchestrate long-running multi-step plans. Released in late 2024 by a small team led by Boris Cherny, it became the reference implementation for "agent that works in your real environment, not a sandbox."
It is the tool to reach for when:
- You want headless agent runs (CI, cron, long refactors).
- You need a real shell, real filesystem, real git — not a webapp sandbox.
- You want to compose subagents, slash commands, and hooks instead of mega-prompts.
Why it stands out
| Feature | What it gives you |
|---|---|
| Native MCP | First-class Model Context Protocol — every MCP server in the ecosystem works |
| Subagent system | .claude/agents/*.md files that Claude dispatches to automatically |
| Slash commands | .claude/commands/*.md with named, parameterized prompts |
| Hooks | Run shell commands on tool-use events (formatters, linters, audit logs) |
| Plugins | A growing marketplace including Anthropic-official plugins like Ralph Wiggum |
| Headless mode | --print, --continue, --resume for CI and scripting |
| 1M context | Opus 4.6 / 4.7 with 1M token window for large codebases |
How to install
# macOS / Linux / WSL
curl -fsSL https://claude.com/install | bash
# Or via npm
npm install -g @anthropic-ai/claude-code
After install, run claude in any project directory. The first run prompts for your Anthropic API key (Pro plan or pay-as-you-go) and writes a .claude/settings.json template.
What to install in Claude Code
The Claude Code surface is intentionally minimal — the value comes from what you put in .claude/. Three packs cover the day one set:
- MCP Server Stack — Chrome DevTools, Postgres, Filesystem, GitHub, and the rest
- Boris Cherny Files — Ralph Wiggum loop + code-simplifier + 5 community subagents
- Awesome Claude Code Subagents — 130 subagent collection + 8 production picks
Install the three with tokrepo install and you have the same baseline most senior Claude Code users start from.
Common pitfalls
- Forgetting
--max-iterationson Ralph Wiggum loops — burn-token risk. Always cap. - MCP scope set to home dir — agent reads
~/.ssh/. Always pass project root. - Mixing
--continuewith edited prompts —--resumeis the safe option when you want to keep the conversation but change instructions. - Hook scripts that hang on stdin — Claude Code waits for them; always close stdin in shell hooks.
Most-installed for Claude Code
10 assets that work with Claude Code
Frequently asked questions
Is Claude Code free?
Claude Code itself is free and open source. You pay for the Claude API tokens it consumes — either via the Anthropic Pro/Team plan (subscription with monthly token quota) or pay-as-you-go API credits. Pricing on claude.com/pricing.
Does Claude Code work on Windows?
Yes via WSL2. Native Windows support shipped in early 2026 but WSL2 remains the smoothest experience because Claude Code uses POSIX shell conventions for hooks and tool calls.
What's the difference between Claude Code and the claude.ai web app?
Claude.ai is a chat interface. Claude Code is an agent — it can read your files, run shell commands, edit code, and chain multi-step actions in your real environment. Claude Code uses the same models (Opus / Sonnet) but with the agent harness wrapped around them.
Can I use Claude Code with another model (Gemini, GPT)?
Officially, no — it ships pointing at Anthropic's API. Community projects let you swap in OpenRouter or local models, but compatibility with Claude Code's tool-use format is tool-specific. The official path is Anthropic models.
Does Claude Code work with monorepos?
Yes. The 1M-context Opus 4.7 build handles repos up to ~500k LOC in a single session. For larger, structure work via subagents that target specific paths, or use the search-specialist subagent for retrieval-then-edit flows.
Browse every tool on the home page
8 tools curated · favicon-fetched logos · type-aware bundles
Back to all tools