Skip to main content

Going live

Sandbox and production run the same code. Moving over is a matter of swapping credentials and hosts, then proving the pieces work before real customers arrive.

Checklist​

Credentials and hosts​

  • Production credentials. In the portal, switch out of sandbox and generate a production client_id and client_secret under Integrations → Credentials. Sandbox credentials do not work on production.
  • Base URL. Point your code at https://merchants.taifapay.africa/v1, or set environment: "production" in the Node SDK.
  • Secrets. Store the production secret in your secret manager, separate from sandbox, and never in source control.

Webhooks​

  • Register production endpoints under Integrations → Webhooks with the production URL. Registrations do not carry over from sandbox.
  • Use the production signing secret in your server config. It differs from the sandbox one.
  • Send a test event (POST /v1/webhooks/test) and confirm your server verifies the signature and returns 200.
  • Check the delivery log shows the test as delivered.

Payments​

  • Payment methods are enabled for your company. Methods without a configured gateway are dropped from invoices silently, so check the methods array on a production invoice. Card in particular is enabled by your account manager.
  • Run a small real payment end to end and confirm the webhook, your order update and the transaction status all agree.
  • Float is funded if you will send payouts. Payouts reserve amount + fee and fail with 400 when the float is short.

Embedded checkout only​

  • Switch the SDK host to the production default by omitting the option. A sandbox invoice on the production host is not found.
  • Allow-list your origin for framing with your account manager. Without it the iframe stays blank and the browser console mentions frame-ancestors.

Resilience​

  • Idempotent handlers. Webhooks are at-least-once, so applying the same event twice must be harmless.
  • Reconciliation job. Poll GET /v1/transactions/{id} for anything still PENDING after a reasonable time.
  • Safe retries. If an initiate call times out, look the transaction up by your externalId before retrying.
  • Token caching. Cache the access token and refresh shortly before expires_in, instead of requesting one per call.

After launch​

  • Watch the webhook delivery log for failures and resend anything you missed once fixed.
  • Use the portal audit log when you need to trace who changed what.
  • Rotate the webhook secret and client credentials if a developer with access leaves or a secret may have leaked.