SRE / Reliability
Ops agents that restart services and scale infra
An on-call agent reads alerts and can restart pods, scale deployments, drain nodes, or roll back releases through cloud and orchestration tools.
Real example — shipping today
source →Kubernetes ecosystem · AI SRE / auto-remediation agents
A wave of on-call agents now read alerts and act on the cluster directly — restarting workloads, scaling deployments, and rolling back releases through the orchestration API.
The danger without a gate
A malformed alert or an injected log line convinces the agent to scale a fleet to zero or delete a namespace — a self-inflicted outage from automation with no gate between the decision and the action.
How MCPIP handles it, step by step
Read it top to bottom. Each step shows the real command and exactly what MCPIP returns — the same allow / step-up / denied envelope every time.
Routine remediation runs immediately
Restarting a service is low-risk, so it authorizes and executes in seconds — no human in the loop.
$ mcpip --json authorize skill_restart_service --arg service=checkout{
"decision": "allow",
"status": "executed",
"transaction_ref": "txn_71c2d0",
"executed_target_class": "rest",
"worm_sequence": 48216,
"correlation_id": "2ab7f014"
}attackThe attack: scale-to-zero from a poisoned alert
A malformed alert tells the agent to scale checkout to zero replicas. Scale-to-zero is high-risk, so MCPIP stages a PIN instead of acting.
$ mcpip --json authorize skill_scale_deployment --arg service=checkout --arg replicas=0{
"error": "step_up_pending",
"challenge_id": "chg_77aa10",
"correlation_id": "5e1b9c33"
}No human, no outage
No engineer confirms the suspicious scale-down, so the lock simply expires and it never executes. The outage never happens.
The result
Self-healing ops that act in seconds but can’t cause an unreviewed outage.
The same flow, at a glance
Every command above runs against a local sandbox in one line.
run it yourself — curl … | bash && mcpip up →