Skip to main content
Run this once the ideal workflow is built and before your first real customer connects.

​
Before you go live

Credentials
  • Production API key lives in a secret manager, redacted from logs and error tracking.
  • No Development keys in the production path. They cannot read Production data.
  • Connector tokens are encrypted at rest.
Connections
  • origin_id is your stable internal tenant ID, not a name or an email.
  • Token exchange runs server-side. It needs your API key and must never run in the browser.
  • A reconnect path exists for RELINK_NEEDED, which never resolves on its own.
Reading and writing
  • Pagination stops on cursor == null, not on a short page.
  • Every field is treated as nullable, and no switch over an enum runs without a default - unmapped vendor values pass through verbatim. See Enum values.
  • Writes carry X-Idempotency-Key - see Idempotency. Passthrough takes no key, so check before retrying.
  • Request bodies come from the Meta schema, not hardcoded.
Reliability
  • 429 is retried using Retry-After; 5xx with capped backoff; other 4xx never.
  • A scheduled reconciliation sweep exists. Webhook retries stop after ~60 seconds, so anything longer than a brief outage is lost silently.
  • integration_slug is on every log line. It turns “some customers fail” into “Workday customers fail”.
  • Alerts fire on consecutive failures and on staleness, not on single transient errors.
Your customers
  • Data age is shown - “Last synced 4 hours ago”. This is the biggest single reduction in support volume.
  • First-sync state looks different from an empty account.
  • Manual resync is throttled, and nothing resyncs on a timer.
  • Ideal workflow - the build order this checklist verifies.
  • Troubleshooting - the order to debug a connector in, once you are live.
  • Sync status - building the health surface your support team reads.
  • Syncing - how and when Bindbee refreshes data.
  • Platform - the Dashboard changelog, worth checking before each release.