Features / Kill switch
An AI agent kill switch that works on the very next call.
When an agent misbehaves, you need it to stop now, not after you find the right server. ACPpotamia's kill switch stops one session or every agent on the team, from the console or the API, and takes effect on the next call worldwide.
How it works.
One session or the whole team
Stop a single task, or every agent at once during an incident.
Immediate everywhere
Enforced by your team's control object, which every call passes through, so there is no cache to wait for.
Webhooks on Enterprise
Get a signed webhook when a session or the team is stopped, to page your on-call or open a ticket.
Stop every agent on the team
curl -X POST https://acp.examopotamia.com/v1/controlplane/kill \
-H "Authorization: Bearer acp_your_admin_key" \
-d '{"reason":"Incident: runaway spend"}'
Questions.
Do calls already streaming get cut off?
Calls already streaming finish; every new call gets HTTP 429 until you revive.
Who can use the kill switch?
Team owners and admins in the console, and keys with the admin scope through the API.
Related features.
Loop detection
Detect and stop AI agents stuck in loops: repeated tool calls, retries and identical prompts. ACPpotamia stops the agent in seconds, before it burns your budget.
Audit logs
Every AI agent call logged: model, tokens, cost, outcome and timing, blocked calls included. Searchable for 365 days, with optional request and response logs.
Budgets and cost control
Put a hard monthly budget on OpenAI and Anthropic spend. ACPpotamia reserves each AI agent call's cost before it is sent, so agents cannot overspend.