Platform
Built to scale
How Orchestrator hands out work to Robot fleets, keeps queues on schedule and shows you what is happening.
Job dispatch
Work reaches the right Robot once
Claims without contention
Robots claim jobs with FOR UPDATE SKIP LOCKED on a purpose-built index, so they do not wait on each other.
Fenced leases
Each attempt holds a versioned lease with a fencing epoch, so a stale Robot cannot renew work it has lost. Results are idempotent.
Runtime-aware placement
Jobs go only to Robots with the runtime they need, and inputs are checked against the entry point schema first.
Recovery sweeps
Expired leases and queue items whose lease was lost are picked up and returned to the flow.
Queues and triggers
Keep work on schedule
Queues with SLAs
Priority, deadline, SLA and risk-SLA per queue. Up to 20 retries for application errors, none for business exceptions.
Cron and queue triggers
Time triggers with time zones and queue triggers, run by a built-in scheduler that fires each time slot once.
Roles that scale apart
One Orchestrator image runs as API, gateway or worker, so each role can be sized for its load.
Robot fleets
Know what every Robot is doing
Presence
A heartbeat every 15 seconds; a Robot is offline after three missed beats. Fleet summaries and filters update live in the browser.
Resilient connections
Exponential backoff with jitter, a local outbox for results and logs, and reconciliation at start-up.
Upgrades that drain
Robots finish running attempts before they upgrade in place.
Measured
Load test results, with their context
Measured on our development topology, not a production cluster: 1,000 connected Robots at 1,000 job starts per minute, with every HTTP call under 200 ms. We publish larger figures only after independent verification.
Observability
See inside the control plane
OpenTelemetry
Traces and metrics over OTLP from the API, HTTP calls, the database and Orchestrator's own meters.
Correlation IDs
Follow one request across logs, traces and audit events.
Health probes
Liveness and readiness endpoints wired for Docker and Kubernetes.
Safe upgrades
The schema only changes additively, and backup and restore are scripted and rehearsed.
Bring your hardest deployment questions.
Talk to an architect