> claude code

Anthropic's terminal coding agent — start and resume sessions, drive it headless from scripts and CI, wire up MCP servers, and keep it on a leash with permissions and CLAUDE.md

getting started sessions cli flags & headless slash commands keyboard shortcuts memory & settings permissions & tools mcp, agents & hooks

// getting started

5 commands

Install Claude Code on your machine

$ curl -fsSL https://claude.ai/install.sh | bash

Start your first session in a project

$ cd your-project

Update Claude Code to the latest version

$ claude update

Diagnose a broken installation

$ claude doctor

Sign in or check which account is active

$ claude auth login

// sessions

5 commands

Continue the most recent conversation

Resume a specific past session

Name a session so it is easy to find later

$ claude -n "payment-refactor"

Run a session in an isolated git worktree

$ claude -w feature-auth

Launch a session in the background and monitor it

$ claude --bg "investigate the flaky test"

// cli flags & headless

7 commands

Get a one-shot answer and exit

$ claude -p "what does the auth module do?"

Pipe command output into Claude

$ cat build-error.txt | claude -p "explain the root cause of this error"

Get structured JSON output for scripts

$ claude -p "summarize this project" --output-format json | jq -r '.result'

Stream responses as real-time JSON events

$ claude -p "explain recursion" --output-format stream-json --verbose --include-partial-messages

Cap turns and spend in automation

$ claude -p "fix the failing tests" --max-turns 5 --max-budget-usd 1.00

Choose or pin the model for a session

$ claude --model sonnet

Give Claude extra directories and prompt instructions

$ claude --add-dir ../shared-lib ../api

// slash commands

7 commands

Reclaim context by compacting the conversation

Start a fresh conversation without restarting

See what is filling the context window

Check token usage and cost for the session

Undo Claude's edits with a checkpoint

Generate a CLAUDE.md for the project

Review a pull request or scan the diff for vulnerabilities

// keyboard shortcuts

8 commands

Interrupt Claude in the middle of a task

Rewind the conversation with a double Esc

Switch permission modes on the fly

Search your command history

See the full transcript with tool details

Run a shell command without leaving the prompt

Reference a file by path in your prompt

@

Edit the prompt with vim keybindings

// memory & settings

5 commands

Give Claude persistent project instructions

$ ~/.claude/CLAUDE.md

Save a fact to memory during a session

Import other files into CLAUDE.md

$ See @README.md for the project overview.

Configure Claude Code with settings.json

$ ~/.claude/settings.json

Set environment variables and a default mode via settings

$ { "env": { "AWS_PROFILE": "dev", "DISABLE_AUTOUPDATER": "1" } }

// permissions & tools

4 commands

Understand and start in a permission mode

$ claude --permission-mode plan

Skip all permission prompts

$ claude --dangerously-skip-permissions

Pre-approve or deny specific tools for a run

$ claude -p "run the tests and fix failures" --allowedTools "Bash,Read,Edit"

Layer persistent allow and deny rules in settings

$ { "permissions": { "allow": ["Bash(npm run test:*)"], "deny": ["Read(./.env)"] } }

// mcp, agents & hooks

6 commands

Add a local MCP server that runs a command

$ claude mcp add my-server -- npx -y my-mcp-server

Connect a remote HTTP MCP server

$ claude mcp add --transport http sentry https://mcp.sentry.dev/mcp

List, inspect, and remove MCP servers across scopes

$ claude mcp list

Create a custom slash command

$ .claude/commands/summarize.md

Create a specialized subagent

$ .claude/agents/code-reviewer.md

Run a shell command on lifecycle events with hooks

$ { "hooks": { "PostToolUse": [ { "matcher": "Write|Edit", "hooks": [ { "type": "command", "command": "npm run lint --silent" } ] } ] } }