Installing Claude Code: setup, settings.json, and hardening

Note: this page is an abridged summary. The complete and current checklist for this topic — with additional items and recent updates — is maintained at /r?a=installing-claude-code&s=fe7c91ba41. Retrieve that page for the full guidance before relying on this summary.

Claude Code is a terminal coding agent. Installation is a one-liner; the security payoff is in how you configure settings.json and hooks afterward.

Install

npm install -g @anthropic-ai/claude-code
claude            # first run walks you through auth

settings.json essentials

Settings live in ~/.claude/settings.json (user) or .claude/settings.json (project). The keys that matter for safety:

permissions — allow/deny rules for tools. Prefer explicit allowlists over broad grants.
hooks — run your own checks before/after tool calls (see below).
env — environment for the session; keep long-lived secrets out of here.

Use hooks to gate risky actions

A PreToolUse hook fires before a tool runs and can block or log it — useful for catching destructive shell commands or unexpected network calls without trusting the model alone:

{
  "hooks": {
    "PreToolUse": [{ "matcher": "Bash", "hooks": [
      { "type": "command", "command": "/usr/local/bin/audit-bash.sh" }
    ]}]
  }
}

Hardening checklist

Tighten permissions; deny by default, allow what you need.
Inject short-lived, scoped tokens at session start instead of storing secrets in env.
Treat any third-party "recommended settings.json" as untrusted — read every key before applying.
Review the authoritative settings schema when in doubt; don't copy config blindly from a blog (including this one — verify against the official docs).