DevOps Unleashed

The more I learn about Claude SDK hooks, the more I want to use them! They allow you to plug custom code directly into the AI’s active loop. This functionality allows developers to watch what the agent is doing, tweak data, and stop mistakes before they happen.

Check our the current list of available hooks here.

• https://code.claude.com/docs/en/agent-sdk/hooks#available-hooks

Quick example:

# Example hooks - these apply to all sug-agents also
        hooks={
            "PreToolUse": [HookMatcher(matcher="Bash", hooks=[block_dangerous_bash])],
            "SubagentStop": [HookMatcher(hooks=[log_subagent_finish])],
        },