Record the browser task you want to automate with Spion, let it reconstruct the steps into structured logic, then export that as an n8n workflow. You import the result into n8n, add credentials, and run it. No node-by-node wiring and no hand-edited JSON.
Why the blank canvas is the real problem
n8n is a powerful tool once a workflow exists. The hard part is rarely the running — it is the building. You know what the task looks like because you do it every week, but translating that lived knowledge into nodes, connections and field mappings takes time and a clear head. Most builders abandon a flow not because n8n is difficult, but because reconstructing the steps from memory is tedious.
The conventional shortcut is to copy someone else's JSON and edit it. That works until the structure does not match your task, at which point you are debugging a stranger's logic. There is a better starting point: capture your own task as it actually happens, and have it turned into a flow you can import.
What you need before you start
- A repeatable browser task — something you do the same way each time, such as pulling figures from a dashboard into a sheet, or copying records between two tools.
- The Spion Chrome extension — free, and the thing that does the recording and reconstruction.
- An n8n instance — cloud or self-hosted. You only need it at the import stage, not while recording.
- Your credentials to hand — API keys or logins for the apps in the flow. Spion captures the steps, not your secrets, so you add credentials in n8n afterwards.
Step one: record the task once
Open the tab where the task begins and start the recording. Then do the task exactly as you normally would — click through the screens, type the values, navigate between tabs. There is no special syntax and nothing to narrate. The goal is to perform the work, not to describe it.
Two small habits make the recording cleaner. First, do the task at a normal pace and avoid stray clicks that are not part of the real flow. Second, run the task end to end without stopping halfway, so the sequence is complete. A clean single pass produces a clearer reconstruction than three hesitant attempts.
The discipline is simple: record the real task, not an idealised version of it. Spion reconstructs what you did, so the more faithful the recording, the more accurate the workflow.
Step two: let Spion reconstruct the steps
When you stop recording, Spion does not hand you a raw clickstream. It reconstructs the task into structured steps — which app, which action, which inputs, in what order. This is the part that usually eats your evening: working out that step three reads a value, step four filters it, and step five writes it somewhere else. You get that structure without building it.
Review the reconstructed steps before exporting. Check that the order matches your intent and that no accidental detour crept in. This is also where you can spot a step that should be conditional or a value that ought to be dynamic rather than fixed. If you want to understand how this capture-first approach differs from older techniques, see workflow discovery vs process mining vs task mining.
Step three: export to n8n
Choose n8n as the export target. Spion produces a workflow that maps your recorded steps onto n8n nodes — triggers, actions and the connections between them — rather than a flat list you have to assemble. The output is a structured flow, not a transcript.
The phrase "no JSON editing" is the point here. You are not opening a file to fix indentation, rename node keys or repair broken connections. The export is shaped to import cleanly, so the manual work that usually follows a copy-paste workflow is removed. If you want the broader picture of how recordings become automations across tools, turning a workflow recording into a Zapier, Make or n8n automation covers the full pattern.
Step four: import into n8n and add credentials
In n8n, create a new workflow and import the exported flow. The nodes and connections appear ready-built. What is deliberately missing is your credentials — Spion records what you did, not your passwords, so each app node will prompt you to attach the right credential.
Add those credentials once and they are reused across runs. For a write-to-sheet step you connect your Google account; for an API call you paste the key. This is a quick, one-off setup and it keeps your secrets where they belong: in n8n, under your control.
| Stage | You do | Spion does |
|---|---|---|
| Record | Perform the task in the browser | Capture every step |
| Reconstruct | Review the steps | Turn clicks into structured logic |
| Export | Pick n8n as the target | Map steps onto nodes and connections |
| Import | Load the flow, add credentials | Produce import-ready output |
| Run | Test and schedule | — |
Step five: test, then schedule
Run the workflow once manually with real inputs and confirm the output matches what you produced by hand. This is the single most useful check — it catches anything the reconstruction read differently from your intent. Compare the result against a recent manual run of the same task.
Once a run is clean, attach the trigger you want. A schedule node turns a weekly report into something that fires every Monday morning; a webhook lets another system kick it off. From here the flow behaves like any other n8n workflow, because it is one — the only difference is how you built it.
Common adjustments after import
- Dynamic values — if you recorded a fixed date range, swap it for an expression so the flow reflects the current period each run.
- Error handling — add a retry or an error branch on nodes that call external services, so a transient failure does not stop the whole flow.
- Filtering and conditions — if your task only acts on certain records, confirm the IF or filter logic matches the rule you apply by eye.
- Notifications — bolt on a Slack or email node at the end so you know each run completed.
None of these require rewriting the flow. They are the normal refinements any n8n builder makes, layered onto a structure you did not have to assemble from scratch.
When recording beats building by hand
Recording shines for tasks you already perform repeatedly in a browser: report assembly, data transfer between tools, onboarding steps, recurring lookups. If you can demonstrate it, you can record it. For tasks that live entirely in code or never touch a browser, the canvas-first approach still has its place. The judgement is simple — if the task is something you do, record it; if it is something you would have to invent from documentation, build it.
For more examples of tasks that suit this approach, browse the use cases, or read how to automate repetitive browser tasks without code for the wider context.
How Spion fits
The hard part of automation has always been capturing the workflow accurately. n8n gives you somewhere to run it; Spion gives you a faithful starting point without the blank canvas. You record the browser task once, Spion reconstructs it into structured logic, and you export a ready-to-run n8n workflow — no manual node wiring and no hand-edited JSON. Add your credentials, test a run, attach a trigger, and the flow is live. Spion is a free Chrome extension, and the same recording can also become a Zapier, Make or n8n automation, or a step-by-step guide, depending on what you need next.