Connector live proof
Från riktig credential till replay-bevis
Den här ytan är sista integrationsbeviset: inte bara att frameworket finns, utan att en riktig provider kan kopplas, synkas, replayas och signeras av.
Decision
Ready for first authorized tenant
Framework, diagnostics and recovery are in place. This lane turns green only after a real provider credential and replay run are archived.
5
proof phases
5
connector families
99%
replay target
Proof phases
Order to run for the first customer connector
Credential intake
Proof: Tenant-approved credential or sandbox credential, provider, scope, expiry and owner are registered.
Pass: Credential can be tested without exposing raw secret values in UI or artifacts.
Provider health check
Proof: Setup endpoint, auth check, provider latency, throttling state and failure reason are archived.
Pass: Provider status is customer-safe and operator-actionable.
Happy-path sync
Proof: Create/update event reaches provider and provider ID is stored on Flow object.
Pass: Flow can reconcile internal ID to external ID without manual lookup.
Replay and recovery
Proof: Webhook replay or retry path is executed for one known event and one failed/recovered event.
Pass: Retry outcome, attempt count, final status and customer-safe explanation are visible.
SLA sign-off
Proof: Delivery certainty, queue backlog, replay success and known limitations are signed off.
Pass: Connector can be sold as live-supported for the selected provider and tenant.
Provider order
Start with highest sales value
- CRM: HubSpot · lead/contact create, update replay, dedupe
- Helpdesk: Zendesk or Freshdesk · ticket create, comment sync, status update, webhook replay
- PMS: Mews or Apaleo · block create, update, release/rollback, provider block ID
- Calendar: Google Calendar or Outlook · conflict check, create, reschedule, cancel
- Messaging: Twilio · dispatch, delivery receipt, failed reason, fallback retry
Guardrails
Credential-safe proof
- Do not paste provider secrets into screenshots, tickets or exported proof packs.
- Every live-proof run must include tenant id, provider id, owner, timestamp and environment.
- A provider can be marked live-supported only for the exact event types that passed.
- Failed replay is acceptable only when the recovery reason, owner and retest state are archived.
- Customer-facing connector status must explain degradation without exposing raw logs.