Ops Integrations

Connectors pull signals into Eridian Ops and, optionally, write back acknowledgements or Slack threads. Outbound writes that change production paging state are high_risk on Enterprise.

Create a connector

POST /v1/ops/connectors
Authorization: Bearer eridian_sk_...

{
  "project_id": "prj_ops_001",
  "type": "pagerduty",
  "name": "pd-payments",
  "credentials_ref": "sec_pd_payments",
  "region": "eu-west-1"
}

credentials_ref is a secret already stored in the project. The API never returns the token. Required scope: ops:write plus governance:manage to create the secret.

PagerDuty

Map PD services to catalog services with pagerduty_service. Unmapped PD incidents still open with service_id: null.

Slack

{
  "type": "slack",
  "channel": "C012345",
  "events": ["incident.opened", "ops.postmortem.published"]
}

Slack delivery is a webhook under the hood. Verify nothing in Slack that you would not put in a ticket. PII redaction applies to assistant answers posted via Slack if you use the Eridian Slack app. Custom bots are your problem.

Datadog / OpenTelemetry

Metrics, logs, and traces are read-only. You configure a query identity. Eridian scans in-region. For OTel, point a collector exporter at the Eridian ingest endpoint issued per connector (https://ingest.geteridian.com/v1/otel with a connector token). Ingest host is not api.geteridian.com.

ServiceNow

Incidents can create or update a SNOW incident. Mapping is field-level. Close in Eridian does not close SNOW unless writeback: true.

Health

GET /v1/ops/connectors/cn_pd_payments/health

{
  "status": "healthy",
  "last_event_at": 1718400000,
  "lag_seconds": 4
}

Lag over 120s on PagerDuty is a SEV-2 for the Ops project. See platform Runbooks if Eridian ingest is the cause.

Disable

POST /v1/ops/connectors/cn_pd_payments/disable

In-flight queries fail with 403 connector_disabled. Open incidents remain.

See Ops Webhooks for push without a vendor app, and Webhook Signing.