Guardian icon

Guardian

Policy-based safety gate for AI agents and workflows. Check actions against Guardian policies, enforce previously-evaluated intents, or gate AI agent tool calls. Returns Allowed (including OBSERVED advisory decisions), Denied, or Needs Approval.

Actions3

Check → Evaluate & Confirm

AI-generated

Summary

Evaluates an action against Guardian policies and, when the decision is ALLOW, confirms execution by submitting the payload to Guardian. Routes results to Allowed, Denied, or Needs Approval outputs; the node records the confirmation but does not perform the downstream action itself.

Inputs

  • Action Type (required) — Action identifier that must match a Guardian policy, such as payment.send.
  • Payload (required) — JSON data evaluated by the policy. Invalid JSON causes an error rather than being treated as an empty object.
  • Requester — Optional name recorded for the audit trail.
  • Project Slug — Optional Guardian project to scope the policy check; empty uses the default project.
  • Idempotency Key — Optional key sent to Guardian to prevent duplicate processing on retries.
  • Test Mode — When enabled, sends Guardian’s test-mode header; the UI states this does not consume quota.

Output shape

Three output branches: Allowed, Denied, and Needs Approval. Each routed item includes Guardian’s response and an _guardian object with the decision, intent run ID, action type, operation, and execution status.

OBSERVED decisions are routed to Allowed as advisory results and are not confirmed as executed. ALLOW triggers a separate execution-confirmation request; duplicate execution may be blocked and routed to Denied, except when the claim belongs to the same workflow execution. DENY routes to Denied, and other decisions such as REQUIRE_APPROVAL route to Needs Approval. The node confirms the intent with Guardian; it does not itself carry out the action.

Links

Discussion