An emergency mechanism that immediately stops an AI agent from taking any further actions when triggered by an operator.
A kill switch is the emergency stop button for your AI agent. When activated, the agent immediately stops processing messages, executing tools, and taking actions. All in-flight operations are terminated.
You need a kill switch because AI agents can enter failure modes that are hard to predict: runaway loops, cascading errors, prompt injection attacks, or simply taking actions that cause harm. The kill switch gives you instant control.
In OpenClaw with Clawctl, the kill switch is accessible from the dashboard, CLI, and API. It works across all channels simultaneously — one click stops the agent everywhere.
Without a kill switch, stopping a misbehaving agent requires SSH-ing into servers, finding the right container, and manually stopping it. In an emergency, that takes too long. A kill switch provides instant, reliable shutdown.
Clawctl provides one-click agent suspension from the dashboard. The kill switch stops all agent activity across all channels instantly. Auto-recovery is paused until you explicitly restart the agent.
Try Clawctl — 60 Second DeployImmediate. The agent stops processing within seconds of activation.
Yes. The kill switch stops the agent on WhatsApp, Slack, Discord, Telegram, and all other connected channels simultaneously.
In-progress tool calls are terminated. Pending messages receive a "service unavailable" response.
Agent Suspension
Temporarily disabling an AI agent so it stops processing messages and executing actions, without destroying its configuration or data.
AI Guardrails
Safety boundaries that constrain what an AI agent can and cannot do, preventing harmful or unintended actions.
Agent Monitoring
Real-time observation of AI agent behavior, performance, and health — including conversation quality, error rates, and resource usage.
Approval Workflow
A process where risky agent actions are paused and routed to a human for review before execution.