Connections

Which of your systems the platform is allowed to read. Any member can read this page. Every action here that changes or re-checks saved configuration, on a connection or on the Slack connection and its channel subscriptions, is reserved to a workspace owner or admin, and is refused during a platform support session.

Your connections

The page opens on saved connections. Search by name, provider, or scope, filter by provider, or select Needs attention to find unfinished setup and reported failures. Access and Data activity are separate: a successful access check does not prove that polling or events work.

Open a connection to review its evidence and configuration. Its address can be bookmarked or shared with another workspace member. Back to connections returns to your filtered inventory.

Add connection

Add connection opens the searchable provider catalog: evidence providers and Slack, filtered by category, each card showing how many connections you have saved. The click path, the shape every wizard shares, and what Review and Verify do are in Set up a connector; each provider's walkthrough is a page in Connect your tools.

Access and activity

Each saved source shows its scope and the latest recorded evidence. Inventory labels include:

Badge	Means
Verified	Access was proved at a point in time, and the source is switched on
Not verified	Saved, but never checked
Verification failed	The last check could not read what it needs
Credential missing	Configured, but with no usable stored credential
Disabled	Saved and switched off
Events not configured	Optional event delivery was left unset
Awaiting first event	Delivery is configured but no successful event is recorded
Event delivery failed	The latest event delivery reported a failure
Polling failed	Polling or a project's snapshot reported a failure
Snapshot received	A successful snapshot was recorded
On demand	Reads run during investigations rather than continuous polling

Verification is a point-in-time proof, not a promise. A token revoked an hour after it was checked still shows as verified until the next check, which is why polling status and event delivery are shown next to it: together they tell you whether the source is still producing usable data.

Actions in connection details

Action	Does
Manage	Reopens the setup wizard with your saved settings, to change scope or replace the credential
Retest	Re-checks access now and updates the status
More actions → Disconnect	Deletes the configuration and the credential, after a confirmation

Slack details use Edit Slack for credentials; Inbound focuses on channel subscriptions and message processing. How to change scope, rotate a credential, and disconnect is in Set up a connector.

Connector against data source

A connector is the integration; a data source is one configured instance of it. The distinction is explained in Connect your tools, and what every connector can and cannot do in What every connector has in common. For how connectors reach your network safely, see Network policy.

Alert lifecycle and existing incidents

Each connection declares alert lifecycle coverage independently of investigation tools or generic repository events. Evidence-only and incomplete connectors cannot authorize automatic recovery. Unknown notifications remain visible, with strict recovery policy and an unresolved signal state. AI may suggest a recovery, but wording and model output do not change provider lifecycle.

For an existing incident, open its connection's Alert lifecycle coverage panel:

1. Select the open incident by title, then its provider signal by summary. The list holds up to the 100 highest-priority open incidents and says when more are open.
2. Enter the exact provider monitor ID and, for Datadog, the exact group scope.
3. Preview the provider episode and check its timing and current result.
4. Enter a reason and bind the verified episode. A concurrent signal, lifecycle, or connector change invalidates the preview and requires a fresh check.
5. Select Provider signals clear in the incident's existing resolution policy control, then reconcile the connection. Original notification evidence remains in the audit trail.

StatusCake uptime and exact Datadog group reads support bounded reconciliation. Native Alertmanager and Grafana recovery uses authenticated event replay. Unsupported families, missing permissions, missing history, conflicts and stale generations produce a visible reason instead of inferred health. Existing regex-derived provider labels alone do not establish verified recovery.

Route the provider webhook paths directly to the API behind your reverse proxy, without an interactive browser-login challenge. Provider authentication remains required. The application supports /webhooks/alertmanager/, /webhooks/datadog/, /webhooks/grafana/, and the per-test StatusCake endpoints the platform creates; deployment-specific proxy allowlists must include the routes you enable.

Current provider verification and accepted recovery evidence are separate. A verified firing read shows the current provider state while preserving the historical notification. It does not authorize closure. The additive provider_clear_generation signal field records only an accepted exact provider clear under the current connector generation. Existing records start without this proof; reconciliation must obtain fresh exact recovery evidence before queued automatic closure can proceed. Changing credentials or connector settings requires re-verification under the new generation.