Four surfaces — the canvas, runs, connections, docs — over one execution engine that treats reliability as a feature, not a plan tier.
Downstream runs once with the whole array — n8n style. Right for rollups, digests, bulk writes.
Downstream runs once per item — Make style. Right for per-lead texts, per-row updates.
Set it per workflow. Override it per node. No other tool gives you both.
A step fails. It retries — twice, with a wait between. Only then does the catch branch take over. You configured all of it in two fields.
Max tries and wait-between, set on any node.
Route failures down their own branch.
Guard a whole branch; retry before the catch runs.
Per-step payloads, timing, and errors — every run.
Re-drive a run with pinned or modified data.
One-click rollback to any prior version.
Signed test + production URLs, custom responses.
Recurrence rules, down to the minute.
Fire from the workflow page.
First-party Opscend events — no plumbing.
Plus a 38-service credential catalog and the HTTP escape hatch for everything else.
Claude-backed — plain text or forced-JSON replies, straight into the flow.
Describe the step in English. The JavaScript is written fresh each run, shaped to the live input.
Scaffold or modify a workflow by describing it, right in the editor.