n8n's canvas shows you the run — it just can't let you change it. Here's the honest path over, including the parts that don't transfer.
In n8n, an execution runs the published snapshot of the workflow. Your edits are drafts until the next publish, and once a run starts its route is fixed — you can watch each node light up, but the only mid-run lever is stop. If a hop is heading to the wrong place, you stop the run, edit, publish, and start over.
On LiveGraph the canvas stays live while a run executes. Every hop is visible as it happens, a run parks at a human approval gate if you want one, and dragging an edge mid-run steers the run you're watching — the drag changes where this run goes next; it doesn't rewrite the saved graph unless you choose to keep it.
We don't offer an n8n file importer — our importer covers Flowise exports only. For everything else the path is generate, not import:
| n8n | LiveGraph |
|---|---|
| Schedule / Webhook / Trigger nodes | Runs start from schedules, webhooks, chat, or the canvas Run button |
| AI Agent / LLM Chain nodes | Worker nodes — role, prompt, and model per node; any provider, not one vendor |
| IF / Switch nodes | explicit edges, or auto edges where a router picks the next specialist |
| Wait / human-approval steps | Approval gates — the run parks at /approvals with the pending payload visible |
| HTTP Request / Code nodes | Per-node tools — http_request, file access, or MCP servers with an exact-name allowlist |
| Error workflows / Error Trigger | A reviewer or fallback node routed like any other edge — visible on the canvas, not a separate hidden flow |
http_request endpoints, or file access — wired per node, a few minutes each, but wired by hand.{{ }} templating has no counterpart — LiveGraph routes on edges and prompts, and a node's job is described in words rather than per-field bindings.If a workflow is pure deterministic plumbing — record in, transform, record out, no judgment calls — n8n's connector catalog is genuinely better for it and moving buys you nothing. LiveGraph earns the move on the runs where a hop can go wrong and a person is on the hook: multi-agent routing, approvals, and the ability to redirect a live run instead of restarting it. Move those first.
Create an account, paste your simplest workflow's export into the describe box, and run it once — you'll know from a single live run whether the canvas earns the rest. Or watch the live demo first: a real multi-agent graph, mid-run edge drag included, no signup.