Google Calendar

A1Connection architecture

The connection uses OAuth 2.0 against the vendor's own consent screen. The authorisation request names the minimum scopes the features need — the exact scope list is shown in the security-flow panel below, pulled from the same provider registry the application uses. The code-for-token exchange happens entirely server-side (client credentials in the token request body, per the vendor's token endpoint contract); tokens are stored encrypted at rest and are never rendered back to any screen — connection pages show presence, not values.

Token lifecycle is handled at a single chokepoint: expiry triggers an automatic refresh, rotated refresh tokens are persisted, and a refresh that the vendor rejects surfaces as a visible reconnect prompt — never as silently broken features. Revocation works from either side: disconnect here, or revoke in the vendor's own security settings.

A2Write semantics and data flow

Every data movement is an explicit action with a logged result. Writes happen on your click — or automatically only where you enabled a rule (auto-push on hire is off by default, per-provider). Reads — imports of people, accounts, or files — run when you press Import, deduplicate against what you already have (clients by name, people by email), skip rather than overwrite, and report created-versus-skipped honestly, which is why re-running any import is safe by design.

Each action writes a row to the app-activity journal (ats_app_activity): what ran, when, for which record, and the outcome — including the vendor's own error text verbatim when something fails. Usage reporting inside the ATS aggregates that same journal, so integration reporting and integration reality cannot diverge.

Anything that leaves the request path — notification fan-out, webhook delivery, activity journalling, mail — runs in fire-and-forget background threads. A slow external endpoint can never make the interface hang, and a failed side effect is logged rather than silently retried into inconsistency.

A3Operational considerations

Connections are organisation-level and gated to owner and admin roles; recruiters use the features a connection powers but cannot connect, disconnect, or reconfigure. Disconnecting removes stored credentials immediately and stops the dependent features visibly, not silently. Data already imported stays yours and editable.

Imported people arrive marked as imported with conservative privacy defaults — no consent is assumed for anyone who never filled in your application form, and retention defaults apply. Everything written is yours to take: CSV exports and the Data Export app cover the same stores the product itself reads. The exit is as open as the entrance — by design, not concession.

A4Placement in the integration topology

This integration is live in the registry today. One connection per provider unlocks every feature it powers, and the topology grid below shows the neighbouring connectors in the same capability area — statuses come from the same registry that drives the in-app hub, so this page can never claim more than the product does. For anything the catalogue does not cover, Webhooks and Zapier are the generic, documented escape hatch.