SafeAgent Execution Guard icon

SafeAgent Execution Guard

Durable execution-claim guard backed by the free SafeAgent test API (POST /claim/test - limited to 10 calls per IP address, total). Claims an (Agent ID, Action Type, Scope) slot before running a side-effectful action, then routes to Proceed (new) or Skip (duplicate). Call Settle after the action completes to record the result. PENDING means the external outcome is unresolved and requires reconciliation. For unlimited, paid production usage, call SafeAgent's POST /claim endpoint directly outside n8n (see github.com/azender1/SafeAgent) - this node only wires up the free tier so the package has no payment/wallet runtime dependency.

Actions2

Claim

AI-generated

Summary

Claims an execution slot for a specific agent action and scope via the SafeAgent free test API, routing items to Proceed if the slot is available or Skip if it already exists.

Inputs

  • baseUrl — SafeAgent API base URL. Defaults to the production Railway instance.
  • agentId (required) — Identifier for the agent or workflow performing the action.
  • actionType (required) — Short label for the side-effectful action being guarded.
  • scope (required) — Unique context for the execution (e.g., customer ID, order ID). Combined with Agent ID and Action Type to form a request ID.

Output shape

Two parallel branches: 'Proceed' (new slot claimed) and 'Skip' (duplicate or PENDING). Each branch returns the API response JSON.

The response JSON includes the request_id and settlement_token, which must be passed to the subsequent 'Settle' operation. The free test endpoint is limited to 10 calls per IP address, total.

Examples

Example 1: Guarding a payment action for a specific customer.

Agent ID: 'payment-bot', Action Type: 'send_payment', Scope: 'customer:123'.

Links

Discussion