Workflow API
Event-driven execution
curl -X POST https://robo-meister.com/api/v1/workflows/holiday-request/runs.json \
-H "Authorization: Bearer $ROBO_TOKEN" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: holiday-request-2026-0042" \
-d '{
"environment": "sandbox",
"schema_version": "1.0",
"webhook_url": "https://example.test/webhooks/robo",
"input": {"employee_id": "EMP-0042", "start_date": "2026-08-17", "end_date": "2026-08-21"}
}'
The API validates the versioned input, organisation, workspace, actor, installed HR capability and environment before reserving credits. It then creates the canonical HR leave request and dispatches its existing business event into the Flow runtime. Replaying the same idempotency key returns the same run.
HTTP/1.1 202 Accepted
Location: /api/v1/workflow-runs/wrun_0198...json
Idempotency-Replayed: false
{
"id": "wrun_0198...",
"workflow_id": "holiday-request",
"workflow_version": "1.0",
"status": "queued",
"environment": "sandbox",
"created_at": "2026-07-30T12:00:00+00:00",
"links": {
"self": "/api/v1/workflow-runs/wrun_0198...json",
"events": "/api/v1/workflow-runs/wrun_0198.../events.json"
}
}
Queued runs can be cancelled with POST /api/v1/workflow-runs/{runId}/cancel.json. Run execution and webhook delivery use retry-safe workers and the configured dead-letter transport.
Controlled sandbox pilot only. Production workflow-execution claims remain blocked until the supplier-onboarding evidence gate proves discovery, missing-document handling, human ownership and decision, risk, evidence, replay, Command Center correlation, and wallet reconciliation in one synthetic run.
Work often already moves through email, chat, spreadsheets, specialist applications and informal
decisions. The real problem is that ownership, context, requirements, exceptions and evidence are split
between them.
Without coordination
- • Requests arrive through several channels and lose a clear owner.
- • Approvals stall because the next decision depends on hidden context.
- • Missing information is discovered only after work has already started.
- • Exceptions are handled manually without a reusable resolution path.
- • Managers see delay or risk after escalation, not while it can still be prevented.
With coordinated execution
- • Work is connected to a goal, case, owner and measurable outcome.
- • Requirements and missing data are checked before dependent work continues.
- • Routine actions can run automatically while judgment stays with authorised people.
- • Exceptions follow configured escalation, fallback or correction paths.
- • Status, evidence, bottlenecks and outcome metrics remain visible in one operational view.