Most scoping calls reach the same question: can we just build it in n8n? For production work, my answer is usually no. Here's the full argument: the cost figures, the trade-offs, and the case where no-code is the right call.
Take a workflow that fires 100 times a day: a morning ops brief, a lead-routing pipeline, or AI-assisted inbox triage. The figures below compare the same workload on each approach, from a recent client engagement.
The 10x cost gap is the obvious signal. The lock-in underneath it compounds over time, and that's the part worth examining.
Workflow logic should live in your repo, not in someone else's UI.
When you stop paying n8n, the workflow stops with it. The logic never existed outside their UI, the credentials live in their vault, and the pricing can change without notice.
When you stop paying me, the workflow keeps running. It sits in the client's repo, on the client's Vercel account, hitting the client's APIs. The only dependency is on services the client already pays for.
Every change is a commit, with a diff, an author, a timestamp, a test run, and a deploy preview. Reverting a bad change is a single command.
The same change in n8n happens inside a UI on the vendor's server. No diff, no review, no rollback unless someone exported the file before clicking save.
When a code-first workflow fails, you have structured Vercel logs, error traces, and a request body you can replay.
When an n8n workflow fails, the UI reports "node failed." No stack trace, no response body, no local reproduction. On the same error, the debugging time isn't close.
n8n leads with its 500+ integrations. A given engagement uses 3 to 5 of them. The other 495 are weight you carry in the pricing model and never use.
For the ones that matter, the code is short:
Paying $20/mo for a visual editor wrapped around code that fits in a single file never pays itself back.
Code-first has real trade-offs:
For one workflow, a non-technical team, and no plan for a second, use Zapier or n8n. The cost and lock-in concerns don't carry much weight at that scale, and the platform makes a one-off automation possible without an engineer.
For three or more workflows, a steady cadence of new ones, or any case where running cost matters, build the code-first version. That could be me, or any developer who holds the same view. Each workflow after the first adds close to nothing.
Total platform cost for a 3-workflow business runs $73 / year plus API tokens. Compare that to your current SaaS bill.
The platform that runs your business should be one you already own.
Weighing a move from a no-code platform to owned code? The async audit at /workflow-audit/ returns a written assessment within 48 hours.