Claude Code Hooks

Hooks are a way to deterministically hook into Claude’s lifecycle — running your own code at specific points during a session.

Source: Claude Code creator Boris shares 12 ways that most people don’t know about

Getting Started

Ask Claude to add a hook for you:

> Add a hook that logs every bash command to a file

Claude will configure the hook in your settings.json.

What You Can Do With Hooks

Route Permission Requests

Automatically send permission prompts to Slack or another Claude (like Opus) for approval, instead of manually clicking through them.

> Add a hook that routes permission requests to our #claude-approvals Slack channel

Nudge Claude to Keep Going

When Claude reaches the end of a turn and stops, a hook can kick off an agent or use a prompt to decide whether Claude should continue working.

> Add a hook that checks if the task is complete and nudges Claude to keep going if not

Pre-process or Post-process Tool Calls

Intercept tool calls before or after they execute — useful for adding custom logging, validation, or transformation.

> Add a hook that logs all file edits to an audit trail

How Hooks Work

Hooks are shell commands configured in settings.json that execute in response to Claude lifecycle events:

{
  "hooks": {
    "onToolCall": {
      "command": "~/scripts/log-tool-call.sh"
    },
    "onPermissionRequest": {
      "command": "~/scripts/route-to-slack.sh"
    },
    "onTurnEnd": {
      "command": "~/scripts/check-completion.sh"
    }
  }
}

Hooks vs Permissions vs Sandbox

Hooks don’t replace permissions or sandboxing — they add a programmable layer on top. You might use permissions to block rm -rf, sandbox to restrict file access, and hooks to log everything that happens.

How LearnAI Team Could Use This

• Use hooks to log Claude Code activity, route approval requests, and enforce project checks during team development sessions.
• Set up PreToolUse hooks for safety checks on destructive operations in shared repos.
• Use Stop hooks for automatic context handoff and session logging.

Real-World Use Cases

• Route risky permission requests to a team approval channel.
• Log file edits and shell commands for audit trails.
• Run validation scripts after tool calls.
• Trigger continuation checks when a Claude Code session stops early.

Why This Matters

Hooks make Claude Code programmable at the workflow level. Instead of just configuring what Claude can and can’t do, you can build automated workflows around Claude’s behavior — approval flows, audit trails, auto-continuation, and custom integrations with your team’s tools.