n8n vs Make After the Agent Wave: Buy a Workflow, Not a Personality
General agents are having a moment. Most “agent” jobs in a small company are still a workflow with an AI step. That is still n8n or Make territory.
This page is periodically reviewed to reflect current pricing and plan changes.
Fastest win
If you can storyboard the steps on a whiteboard, buy a workflow product. If you cannot, a general agent will not invent a process you would accept. It will invent a process you have to clean up.
Headlines versus jobs
Cowork, Computer, Super Agent and Work will keep winning screenshots. Repeatable jobs still win on explicit steps: trigger, retrieve, transform, AI draft, human approve, write back.
That pattern is older than this news cycle. It is also cheaper to debug. When a general agent fails, you get a narrative. When a workflow fails, you get a node.
n8n when you want control; Make when you want visual speed
That split has not changed because Cowork launched. n8n fits developers and self-hosting. Make fits teams who will not look at JSON. Both beat an unbounded super-agent for repeatable work.
Cost per successful workflow
Count executions or credits, failed runs, and the weekly maintenance. Self-hosted n8n trades vendor margin for your time. Make trades money for less engineering.
Do not compare either to Manus on “intelligence”. Compare them on the same lead-to-report job, including the approval step.
When a general agent still wins
Use a general agent when the job is not repeatable: a messy one-off research-and-assemble task, a folder that will never look the same twice, a prototype you will throw away.
Do not use a 40-node workflow for that. You will spend more time maintaining the workflow than the task is worth.
The error is treating every headline agent as a workflow replacement, or every workflow as capable of messy work. They are complementary.
Ranked recommendation
Best choice for a technical team: n8n.
Best choice for a non-technical ops team: Make.
Best choice for a one-off messy brief: a general agent with a cap, not a 40-node workflow you will not maintain.
Avoid rebuilding your company in a super-agent because workflows feel unfashionable this week.
Key Takeaways
- →Repeatable work wants a workflow product.
- →n8n vs Make is control versus visual speed.
- →Successful workflows are the denominator.
- →General agents are for the non-repeatable remainder.
Editorial context
Who is this for?
Small teams automating lead-to-report or ticket-to-slack jobs.
When NOT to use this
People who need a desktop folder agent once a month. That is Cowork/ChatGPT, not n8n.
Pricing insights
Executions and credits are still the native units. Convert them to successful workflows. Self-hosting is not free; it moves cost into ops.
Alternatives to consider
Taskade if you want bundled agents and a lighter suite. Browse AI if the job is website-to-spreadsheet. A general agent only for non-repeatable tasks.
Final verdict
Repeatable multi-step work belongs in n8n or Make. Save general agents for the messy remainder.
Frequently Asked Questions
Did Cowork make n8n and Make obsolete?
No. Repeatable jobs still want a workflow with an AI step. General agents are for messy one-offs you will not maintain as a 40-node graph.
n8n or Make?
n8n if you want control and can self-host. Make if the team will not look at JSON. Convert executions to successful workflows, including the approval step.
When should we still buy a general agent?
When the job is not repeatable: a folder that will never look the same twice, a prototype, a one-off brief. Cap it. Do not rebuild the company in a super-agent.