Claude Code

Claude Code is Anthropic's agentic coding tool. It runs in your terminal (and VS Code, JetBrains, the desktop app, web, Slack), reads the codebase, edits files, runs commands, runs tests, and creates git commits and PRs.

16.1 Install and start

# macOS / Linux / WSL (native installer)
curl -fsSL https://claude.ai/install.sh | bash
# Windows PowerShell
irm https://claude.ai/install.ps1 | iex
# or via npm
npm install -g @anthropic-ai/claude-code

cd my-project
claude                      # interactive session
claude "fix this bug"  # start with a prompt
claude -p "give a summary" --output-format json   # headless / scripting
claude -c                   # continue the last conversation
claude -r                   # resume an old session (picker)

16.2 How it works

16.3 Memory: CLAUDE.md

The /init command scans the codebase and creates a starter CLAUDE.md. In CLAUDE.md you can import other files with @path/to/file. Keep it short and specific — it is loaded in every session.

# Project: Billing API (.NET 8)

## Commands
- Build: `dotnet build`
- Test: `dotnet test --filter Category!=Integration`

## Conventions
- Async method names end with `Async`.
- Never hand-edit EF Core migrations.
- Update `docs/api.md` on every public API change.

16.4 Slash commands (important)

Custom slash command: create a .claude/commands/fix-issue.md file (with the $ARGUMENTS placeholder) — it runs as /fix-issue 123. In newer versions these have merged with skills.

16.5 Keyboard shortcuts and input tricks

16.6 Permission modes

// .claude/settings.json (project, committed to git)
{
  "permissions": {
    "allow": ["Bash(dotnet test:*)", "Bash(git status)", "Read(./src/**)"],
    "deny":  ["Read(./.env)", "Bash(rm -rf:*)"]
  },
  "env": { "ASPNETCORE_ENVIRONMENT": "Development" }
}

Settings hierarchy (high → low priority): managed policy → command-line args → .claude/settings.local.json → .claude/settings.json → ~/.claude/settings.json.

16.7 Hooks

Hooks are shell commands that run deterministically on Claude Code lifecycle events — not left to the model. For example a formatter after every edit, or blocking a dangerous command.

{
  "hooks": {
    "PostToolUse": [{
      "matcher": "Edit|Write",
      "hooks": [{ "type": "command", "command": "dotnet format --include $CLAUDE_FILE_PATHS" }]
    }]
  }
}

16.8 Subagents

A subagent is a specialized assistant that runs in its own separate context window, with its own tools and system prompt. The main conversation stays clean — the subagent returns only a summary. Create one with /agents or put a file in .claude/agents/.

---
name: test-runner
description: Runs tests after a code change and fixes failures. Use proactively.
tools: Read, Edit, Bash, Grep
model: sonnet
---
You are a test automation expert. Run the tests, if they fail find the root cause,
make a minimal fix, and run them again. Give a short report at the end.

16.9 Plugins, MCP and integrations

16.10 Best workflow

1. Explore: "First read the relevant files, do not write code yet."
2. Plan: Have it make a plan in plan mode (Shift+Tab) and review it.
3. Code: Have it implement, in small steps.
4. Verify: Have it check with tests / build / screenshots (TDD is best).
5. Commit: Have it write the commit message and create the PR.
6. New unrelated task? /clear. Context filling up? /compact.