Automations and SLA

Both are settings: they need an admin user, or an API key with settings:read and settings:write.

Method	Path	Description
GET, POST	/v1/automations	List or create automations
GET, PATCH, DELETE	/v1/automations/{id}
GET, POST	/v1/sla-policies	List or create SLA policies
GET, PATCH, DELETE	/v1/sla-policies/{id}

Objects

type Automation = {
  id: string;
  name: string;
  trigger: "ticket_created" | "customer_replied" | "agent_replied" | "ticket_updated";
  conditions: { field: "subject" | "body" | "from_email" | "channel" | "mailbox_id" | "priority" | "severity" | "status" | "tag"
                     | "sentiment" | "ai_decide" | `field.${string}` | `contact.field.${string}`;
                op: "contains" | "not_contains" | "equals" | "not_equals" | FieldOp;
                value: string;
                question?: string /* ai_decide only: 3..300 characters of plain text */ }[];
  actions: { type: "set_status" | "set_priority" | "set_severity" | "assign_user" | "assign_team" | "add_tag" | "remove_tag"
                 | "send_reply" | "add_note" | "mark_spam" | "ai_triage" | "set_field";
                 value: string;
                 field?: string /* set_field: the ticket field key */ }[];
  enabled: boolean;
  position: number;
  run_count: number;
};
// severity: equals / not_equals with sev1..sev4 or "none". set_severity: sev1..sev4, or "" to clear.
// ai_decide: gt / lt with value a probability 0.01..0.99; at most 3 per automation; all of one
// trigger run are asked in one call (charged once to the AI balance). It does not hold when no decision engine answers.

type SlaPolicy = {
  id: string;
  name: string;
  priority: Priority | null;
  first_response_minutes: number;
  resolution_minutes: number;
  enabled: boolean;
  position: number;
};

Conditions on field.<key> (the ticket’s custom field) and contact.field.<key> (its contact’s) take the operators of the field’s type; equals and not_equals mean eq and ne. set_field writes value as a string: "12.5", "true", "2026-10-11", comma separated option ids, or "" to clear. It is checked when the automation is saved, and skipped at run time if the field was archived or deleted since. add_tag respects exclusive label groups.

See Automations and SLA policies for how they behave.

Create an automation

curl https://api.opendesk.sh/v1/automations \
  -H "Authorization: Bearer od_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Outages are urgent",
    "trigger": "ticket_created",
    "conditions": [{ "field": "subject", "op": "contains", "value": "outage" }],
    "actions": [{ "type": "set_priority", "value": "urgent" }, { "type": "add_tag", "value": "incident" }],
    "enabled": true
  }'

Create an SLA policy

curl https://api.opendesk.sh/v1/sla-policies \
  -H "Authorization: Bearer od_live_..." \
  -H "Content-Type: application/json" \
  -d '{ "name": "Urgent", "priority": "urgent", "first_response_minutes": 30, "resolution_minutes": 240, "enabled": true }'

priority: null applies the policy to any priority.

Events

automation.* and sla_policy.*.