Hooks in Claude Code: Automate Triggers on Commands and Events

What Hooks Are

Hooks are scripts that run automatically at specific points in Claude Code's workflow. They let you automate responses to what Claude Code is doing without modifying Claude Code itself. You can hook into command execution, session start and end, and git operations.

Configuring Hooks

Hooks live in your claude_settings.json:

{
  "hooks": {
    "preCommand": "./scripts/pre-command.sh",
    "postCommit": "python3 ./scripts/notify.py"
  }
}

Each hook runs with the Claude Code context as environment variables — the current working directory, active model, session ID, and the command being executed are all available to the hook script.

Common Use Cases

Pre-command validation: Run linting or tests before Claude Code executes a command. If the checks fail, the hook can exit with a non-zero code to prevent the command from running.

Post-commit notifications: After Claude Code creates a commit, post a Slack message or update a task tracker.

Session logging: Record session metadata to your own logging system for billing or audit purposes.

#!/bin/bash
# pre-command.sh example
echo "Running: $CLAUDE_COMMAND" >> ~/claude-sessions.log

Error Handling

If a hook fails with a non-zero exit code, the behavior depends on the hook type. Pre-command hooks that fail will block the command from running. Post-command hooks that fail log an error but do not affect the command outcome.

Write hooks to be idempotent and fast. A slow pre-command hook will make every Claude Code operation feel sluggish.