Airia has extended its Agent Constraints with Security Runbooks that trigger inline at the moment an AI agent attempts a tool call, the company said on 13 August 2026.
Instead of a simple allow-or-deny decision, the platform runs a customer-defined sequence of steps when a tool call is intercepted. Each runbook returns one of three verdicts: Resume, Block, or Escalate for human review.
Airia said traditional security playbooks follow a detect-and-respond model, so by the time they run the agent has already acted. Moving runbook logic upstream turns it into a prevention mechanism.
Administrators can use if-then-else logic to automate routine sequences while reserving review workflows for high-stakes actions. Every decision is captured in an audit trail showing the checks performed, the conditions evaluated and the verdict returned.
“Running a runbook at the moment of the tool call means you can actually gather what you need to know, confirm the system is in a safe state, and only then let the action through, at the speed agents operate, with the audit trail regulators expect,” said Kevin Kiley, CEO of Airia.
Inline Security Runbooks are available now for Airia customers.