GlobiGuard Governed HTTP Action icon

GlobiGuard Governed HTTP Action

Construct, authorize, recheck, and execute one exact HTTP request in a single fail-closed boundary

GlobiGuard Governed HTTP Action

AI-generated

Summary

Constructs one exact HTTPS request per input item, sends its request summary to GlobiGuard for authorization, then executes it once only when the authorization is current and executable. Routes results to separate outputs for executed, modified, blocked, queued, and error cases.

Inputs

  • Governed Action Type (required) — Semantic policy action name, such as email.send; the HTTP method is authorized separately.
  • Method — HTTP method: DELETE, GET, PATCH, POST, or PUT; defaults to POST.
  • URL (required) — HTTPS destination. URLs with credentials or fragments are rejected, and redirects are disabled.
  • Headers (JSON) — Optional JSON object of string-valued headers. Transport-controlled headers and values containing line breaks are rejected.
  • Body (JSON) — JSON object or array sent with non-GET requests; defaults to the current input JSON. GET requests omit the body.
  • Response Format — Parse the response as JSON or return it as text; defaults to JSON.
  • Timeout (MS) — Downstream request timeout from 1,000 to 120,000 milliseconds; defaults to 30,000.
  • Data Classes — Data classifications supplied to GlobiGuard; defaults to INTERNAL.
  • Purpose — Business purpose used by purpose-limitation policies.
  • Workflow Run ID — Workflow run identifier sent to GlobiGuard; defaults to the n8n execution ID.
  • Correlation ID — Optional correlation identifier sent to GlobiGuard.
  • Idempotency Key — Optional stable business key sent to GlobiGuard. The node does not retry the downstream request.
  • Approved Queue Entry ID — Optional queue approval to consume. GlobiGuard binds it to the current request and issues fresh authorization; the approval itself does not execute the request.

Output shape

Five output branches: executed, modified, blocked, queued, and error.

Each input item is authorized independently. The request's canonical hash and summary are sent for authorization, then the hash is checked again before execution. Only an ALLOW with a valid, short-lived, executable authorization and no unsupported obligations or modifications runs the request. Executed items contain response and GlobiGuard authorization metadata; non-ALLOW decisions are marked NOT_EXECUTED and include decision details. With Continue On Fail enabled, errors go to the error branch with a STOPPED status; otherwise the node throws. Downstream HTTP failures suppress response details.

Links

Discussion