Hook System
Customize your agents' behavior at every key step, with 15+ events available.
Every agent follows a lifecycle: receiving a message, thinking, executing actions, formulating a response. Until now, this cycle was fixed. With the hook system, you can inject your own logic at every step.
A hook is a function automatically triggered when a specific event occurs. Want to log every command executed? Hook on_tool_call. Want to check every response before sending? Hook pre_response.
The system supports 15+ events covering the agent's entire lifecycle. Hooks can be defined in JSON within the agent's configuration, without writing server code.
Declaration
You define your hooks in the agent's configuration: which event, which action, which conditions.
Event
When the event occurs in the agent's lifecycle, the hook is triggered.
Conditions
Optional filters limit execution (e.g., only for messages containing code).
Action
The hook executes its action: API call, context modification, skill trigger, notification.
Chaining
Multiple hooks can listen to the same event. They execute in the defined priority order.
Security audit
An on_tool_call hook records every system action in an audit log. Complete traceability without modifying the agent's code.
Context enrichment
An on_message_received hook queries an internal database and injects relevant information into context.
Notification
A post_response hook sends a Slack alert when the agent produces a response containing a warning or error.
featurePages.hooks.config
- Available for all EasyClaw v2 agents
- No hooks active by default (opt-in)
- Complete documentation for each event available in the toolkit