SoterAIInstall

Summary

Check a planned tool call against an Agent Passport and route allowed items to the Safe output and blocked or approval-required items to the Flagged output. Place this node before the real tool; provide the passport token, session ID, tool name, and action, with optional content, destination, target, and audit metadata.

Inputs

• Passport Token — Passport token returned by Issue Passport; if blank, the node uses the incoming passportToken or passport.token.
• Session ID — Session shared by passport steps; if blank, the node reads it from the incoming item.
• Tool Name (required) — Name of the tool or function the agent plans to call, such as gmail.send.
• Tool Action (required) — Action the tool will perform, such as send_email.
• Tool Content — Optional arguments or payload to inspect; defaults to an empty string.
• Tool Destination — How far the planned call reaches: external, internal, local, or unknown. Defaults to unknown.
• Tool Target — Optional URL, recipient, file, table, or other resource targeted by the call.
• Detection Engine — Choose Auto (cloud first, local fallback), Cloud Only, or Local Only. Cloud checks require a SoterAI credential; local checks use bundled rules with reduced coverage.
• Options — Optional settings include audit metadata JSON, project override, request timeout, batch concurrency, reuse of identical items, raw API response inclusion, and whether Auto mode may fall back to local rules when cloud checks fail.

Output shape

Items are routed to Safe or Flagged outputs.

The notice describes Safe as allowed calls and Flagged as blocked or approval-required calls. With Continue On Fail enabled, failed items also leave through Flagged. Runtime response fields are not established by the provided source excerpt.

Links

• n8n community nodes installation guide