Guides · v1

Verification jobs and safe retries

Verification saves a durable, read-only provider identity check. The worker continues after the HTTP request or chat ends.

Choose an assigned company and a permitted saved connection.
Generate one Idempotency-Key for this exact verification request. Keep it when retrying the same submission.
POST to /v1/workspaces/{workspaceId}/connections/{connectionId}/verify.
Use the returned job ID with the job-detail endpoint to inspect its saved state and result.

http

POST /v1/workspaces/workspace_example/connections/connection_example/verify
Authorization: Bearer <access_token>
Idempotency-Key: verification_example_001
Content-Type: application/json

{}

A queued/running status is not a successful provider response. Preserve unavailable and failed states in your application, and inspect the recorded account/company identity and timestamps before representing a connection as verified.

Retry deliberately

The edge never retries mutations automatically. Retry eligible read requests after a transient failure, honoring Retry-After where present and using bounded exponential backoff with jitter. A verification retry must reuse its original idempotency key and request. Task creation has no idempotency guarantee: after a lost response, inspect current tasks before deciding whether to submit again.

Cancellation and authority

Cancellation requires requester rights or the company’s cancellation grant. Company administrators can read company jobs; other users see their authorized jobs. Worker execution rechecks identity and service rights. A job can stop when those rights change.