The definitive hub for Anthropic's terminal-native agentic coder. Free 5-day course, deep-dive articles, workflow examples, power-user tips, and a head-to-head comparison with Cursor and GitHub Copilot. All free, no signup, no email.
If you have a terminal and an Anthropic API key (or a Claude Pro/Max subscription), you can be running Claude Code in under a minute. Paste the commands below.
# 1. Install (macOS / Linux) curl -fsSL claude.com/install.sh | sh # 2. Authenticate (uses your Claude account) claude login # 3. Jump into any git repo and start cd ~/code/myproject claude "add type hints to src/utils.py and run the tests" # 4. Watch it work. Hit tab to approve commands, esc to interrupt.
A self-paced 5-day course covering installation, multi-file editing, sub-agents with git worktrees, CLAUDE.md configuration, and shipping real projects. No signup, no email, no paywall.
Five structured lessons covering every primitive you need to use Claude Code in production. Each day is ~1 hour of focused reading plus hands-on exercises. By Day 5 you've built a real feature in your own repo using Claude Code end-to-end.
Start Day 1 →Long-form practitioner posts — not vendor marketing. Each one goes under the hood on a specific aspect of Claude Code.
A technical breakdown of terminal access, efficient search, git worktrees, and stopping conditions. With links to every open-source implementation worth reading.
The complete reference guide to Claude Code — features, commands, permission modes, CLAUDE.md patterns, and production workflows.
Head-to-head on real codebases. Which agentic IDE earns the daily driver slot for working engineers in 2026.
Quick-reference card with strengths, weaknesses, use cases, pricing, and alternatives. Cross-linked from every Claude Code mention on the site.
The patterns that show up repeatedly in production use. Copy the commands, adapt to your project.
Spin up three agents working on three features simultaneously. Zero merge conflicts because each worktree is its own physical directory.
git worktree add ../myproj-auth feat/auth git worktree add ../myproj-search feat/search git worktree add ../myproj-429 bugfix/rate-limit # Three parallel Claude Code sessions (cd ../myproj-auth && claude "add OAuth") & (cd ../myproj-search && claude "add fuzzy search") & (cd ../myproj-429 && claude "fix 429 loop") & wait
Drop a CLAUDE.md file at the root of your repo. Claude Code auto-reads it on every run, so you never re-explain project conventions.
# MyProject Stack: Python 3.12, FastAPI, PostgreSQL, SQLAlchemy 2.0 Style: Black formatter, type hints always, Google docstrings # Test commands - Run unit tests: `pytest tests/` - Run integration: `pytest -m integration` # What NOT to touch - migrations/*.py (alembic manages these) - .env.* (secrets, never edit)
Point Claude Code at a failing test and let it iterate until green. It runs the test, reads the failure, edits code, reruns.
$ pytest tests/test_parser.py -x # FAILED tests/test_parser.py::test_unicode_escapes $ claude "fix the failing test in tests/test_parser.py" # Claude reads the test, edits parser.py, re-runs, reports green
Restrict Claude Code to safe commands only. Good for running in a shared repo where you want zero chance of a destructive operation.
# Run in read-only analysis mode claude --permission-mode auto \ "audit src/ for security issues, " \ "report findings, do not edit files"
Have Claude Code apply the same pattern across several repos. A bash loop is all you need.
for repo in api worker frontend; do cd ~/code/$repo claude "rename getUserById to fetchUserById " \ "everywhere. Update tests. Commit." cd - done
Claude Code accepts piped input. Great for error logs, stack traces, and CI failures you want explained or fixed.
$ pytest 2>&1 | claude "explain these test failures " \ "and write a fix plan" $ kubectl logs webapp-7f4b | claude "summarize errors"
Short, punchy, actionable. If you use Claude Code daily, at least three of these will be new to you.
CLAUDE.md at your home directory too. It acts as a global config that applies to every project, not just the current repo. Great for personal preferences like "always use 4-space indent in Python" or "commit messages should end with the Jira ticket."/resume to continue the last session instead of starting fresh. Claude Code keeps conversation history per directory, so you pick up where you left off./clear to reset context when Claude Code gets confused. A stale context window is one of the most common causes of weird behavior — just start fresh and re-explain the task.../myproj-auth-oauth is way easier to tab-complete than ../myproj-exp-1. Small thing, saves minutes every day.--permission-mode acceptEdits when you trust the file changes but want to review shell commands. Lets the agent edit freely but still prompts on anything that touches the shell.pytest 2>&1 | claude "fix these" — Claude reads the full failure context and iterates.claude "analyze this codebase and write a CLAUDE.md file with the key conventions." Then edit to taste./cost with your session to see exactly how many tokens you've burned. Useful on long exploratory sessions where you might be racking up charges.isolation: "worktree" creates a disposable worktree, runs the sub-agent there, and cleans up automatically if nothing changed. Safe way to let the agent experiment.The quick comparison table. Pick the tool that matches your workflow — or run multiple and see which sticks.
| Feature | Claude Code | Cursor | aider | GitHub Copilot |
|---|---|---|---|---|
| Terminal-native | ✓ yes | × GUI | ✓ yes | × IDE |
| Multi-file refactor | ✓ strong | ✓ strong | ✓ strong | ~ limited |
| Git worktree sub-agents | ✓ built-in | × no | × no | × no |
| Project config file | ✓ CLAUDE.md | ✓ .cursorrules | ~ .aider.conf | × no |
| Multi-model support | × Claude only | ✓ Claude/GPT/Gemini | ✓ all major | ~ GitHub-hosted |
| Open source | × closed | × closed | ✓ MIT | × closed |
| Price | Included w/ Pro $20/mo | Free or $20/mo | Free (pay API) | $10/mo individual |
Claude Code is Anthropic's terminal-native agentic coding tool. It runs in your existing shell, uses your existing git setup, reads and edits multi-file projects, runs tests, and can spawn sub-agents via git worktrees. It's included with Claude Pro and Max subscriptions and uses Claude Sonnet 4.6 or Opus 4.6 as its backing model.
Claude Code is CLI-first — it runs in your real terminal and integrates with your real git workflow. Cursor is a VS Code fork with a polished GUI. Claude Code is better for terminal-native workflows, scripting, multi-agent orchestration, and CI pipelines. Cursor is better for visual multi-file editing and IDE-style pair-programming flows. Both are legitimate choices for different workflows.
The CLI binary is free to install. You need either an Anthropic API key (pay-per-token) or a Claude Pro/Max subscription ($20/mo) to actually run it. Most working developers should start with the Pro subscription — it includes a daily quota of agentic coding usage.
CLAUDE.md is a project-level configuration file that Claude Code reads automatically when you invoke it in that directory. It contains project-specific instructions, conventions, allowed commands, and context. Think of it as a persistent system prompt that lives in your repo. You can also put one at your home directory as a global config.
Yes via WSL2. Native Windows support is rougher than macOS/Linux because of terminal rendering and shell quirks. If you're on Windows, install WSL2 and run Claude Code inside an Ubuntu environment — that's the supported path.
Yes. Use --permission-mode bypassPermissions carefully (only in sandboxed CI environments where you control the blast radius) or preset an allowlist in CLAUDE.md. Many teams use Claude Code in GitHub Actions to auto-fix test failures, update dependencies, or generate boilerplate on PR creation.
The 2-day in-person Precision AI Academy bootcamp covers Claude Code, MCP, sub-agent workflows, and production agent engineering. Five U.S. cities. $1,490. June–October 2026.
Reserve Your Seat →