n8n vs Zapier vs Make vs Custom Code for Small Business

All four can move a lead from a web form into a CRM. The difference shows up in month six, when volume grows, the workflow gains AI steps, and the person who built it has moved on.

Quick decision table

What is each tool good at?

Zapier has the largest catalog of integrations and the easiest builder. If your stack is common SaaS (Gmail, HubSpot, Sheets, Slack) and your workflows are "when X happens, do Y and Z," it is the fastest path. It gets expensive and awkward with loops, heavy branching or many steps per run.

Make (formerly Integromat) gives you a visual canvas with routers, iterators and error handlers. It suits ops people who think in flowcharts. Big scenarios get hard to read, and every module operation consumes credits.

n8n is built for technical users. It has a visual editor, but you can drop into JavaScript or Python at any step, call any HTTP API, and run AI agent nodes with tools and memory. The Community Edition runs on your own server, which matters if you care where customer data sits.

Custom code (Python or Node.js with a queue, a database and a scheduler) gives you full control, tests, version control and no per-run fees. You also take on hosting, monitoring and maintenance.

What do Zapier tasks, Make credits and n8n executions mean?

This decides your bill.

• Zapier task: each successful action step is one task; the trigger is free. "New form entry → create CRM contact → Slack message → add sheet row" uses 3 tasks per lead.
• Make credit: consumed per module operation in a scenario, and some AI modules use more. A five-module scenario uses credits for each module that runs.
• n8n execution: one complete workflow run, however many steps. A 3-step and a 30-step workflow each count once per lead.

Per-step billing (Zapier, Make) punishes long workflows. Per-run billing (n8n) does not. For short automations the difference is small; with AI extraction, lookups and loops it adds up fast.

How do 2026 prices compare side by side?

Zapier figures are from Zapier's pricing page as of October 2026, with monthly-billing equivalents from a September 2026 third-party check. We could not verify current n8n and Make prices on their official pages, so we left exact figures out on purpose. Check n8n.io/pricing and make.com/en/pricing before you decide.

Same workflow, four bills: lead form to CRM to WhatsApp alert

Take a common small business workflow:

1. A website form submission arrives (trigger).
2. Check whether the contact already exists in the CRM.
3. Create or update the contact.
4. Ask an AI model to tag the lead (hot, warm, spam).
5. If hot, send a WhatsApp alert to the owner.
6. Log the lead in a Google Sheet.

That is five actions after the trigger. At 300 leads a month:

Meta's WhatsApp fees and the AI model's API cost apply on every option, so they cancel out. At 3,000 leads, Zapier and Make usage multiplies by ten while n8n still counts one execution per lead. That is where many businesses switch.

Is n8n really free? What self-hosting actually costs

The n8n Community Edition is free to run on your own server. "Free" means no licence fee, not zero cost:

• Server: a small virtual server handles most small business workloads.
• Database and backups: beyond testing, run n8n on PostgreSQL and back it up.
• Updates and security: someone applies updates, manages SSL and locks down the editor.
• Monitoring: if a workflow fails at 2 am, who finds out? Set up error workflows and alerts.
• Features: some team and access-control features are paid-only. Compare editions on n8n's site.

If nobody can own those jobs, n8n Cloud or a partner who maintains your instance is cheaper than a weekend lost to a broken server.

When should you use custom code instead of a no-code tool?

Custom code wins when:

• The workflow is your product. If customers pay for the thing the automation does, it belongs in your codebase with tests and version control.
• Volume is high and steady. Tens of thousands of runs a month on per-step pricing gets expensive; a small worker on a server does not.
• Logic is complex. Deep branching, long-running jobs, retries with state, or heavy data transforms are easier to maintain as code than as a 60-node canvas.
• You need strict data control. Regulated data, audit trails, or data residency requirements.
• Real-time responses matter. Voice agents and live chat need sub-second responses that a workflow tool may not deliver reliably.

A common pattern is hybrid: n8n handles orchestration and integrations, and custom code handles the hard parts behind an API that n8n calls.

Can n8n run AI agents?

Yes. n8n's AI agent nodes connect to OpenAI, Anthropic, Google and other models, use tools (APIs, databases, document search) and keep memory. They suit back-office agents: email triage, invoice extraction, drafting replies. For phone calls, run the conversation on a voice platform and let n8n handle what happens after.

Who owns the workflows when the freelancer leaves?

This question costs small businesses more than any pricing tier. Before anyone builds for you:

• Use your accounts. The Zapier, Make or n8n workspace should be in your company's name, billed to your card, with the builder invited as a user.
• Use your API keys. OpenAI, CRM and WhatsApp credentials should belong to you.
• Get exports. n8n workflows export as JSON and can live in a Git repo. Make scenarios can be exported as blueprints. Ask for these at handover.
• Get documentation. One page per workflow: what triggers it, what it touches, what to do when it fails.

If a builder refuses to work in your accounts, treat it as a warning.

Which one when: our verdict

• Pick Zapier if you have fewer than a handful of short workflows, low volume, and nobody technical. Pay for simplicity and move on.
• Pick Make if you want visual branching and data shaping on a small budget, and someone on the team enjoys building flowcharts.
• Pick n8n if your workflows are long, include AI steps, run at growing volume, or touch data you would rather keep on your own server. This is our default for most small business back-office work.
• Pick custom code if the workflow is your product, needs real-time responses, or has outgrown a canvas.
• Combine n8n and code when you need both: n8n for integrations, code for the hard logic.

If you are already on Zapier and it works, do not migrate for its own sake. Migrate when the bill or the workarounds start to hurt.

FAQ

Is n8n really free?

The n8n Community Edition is free to self-host, with no licence fee and no execution limits set by n8n. You still pay for a server, backups and someone's time to keep it updated and secure. n8n Cloud, the hosted version, is a paid subscription priced by monthly executions. Check n8n.io/pricing for current plans, as they change.

Is n8n cheaper than Zapier?

For long or high-volume workflows, usually yes. Zapier charges a task for each action step, while n8n charges one execution per workflow run regardless of steps. A five-step workflow costs five Zapier tasks but one n8n execution. For a handful of short automations, Zapier's simplicity can outweigh the price difference. Compare both pricing pages against your real volumes.

What is the difference between Zapier tasks, Make credits and n8n executions?

A Zapier task is one successful action step; triggers are free. A Make credit is consumed per module operation in a scenario, and some AI modules use more. An n8n execution is one complete workflow run, however many steps it has. Per-step billing makes long workflows expensive on Zapier and Make, while n8n's cost tracks the number of runs.

Should a small business use Make or Zapier?

Choose Zapier if the workflows are simple and the person building them is non-technical, because it is the easiest to learn and has the most integrations. Choose Make if you need branching, loops or data transformation and want lower cost per operation. Make's visual canvas suits people who think in flowcharts. Check both pricing pages against your monthly volume first.

When should I use custom code instead of a no-code tool?

Use custom code when the workflow is part of your product, runs at high steady volume, has complex logic or long-running jobs, handles regulated data, or needs real-time responses like voice. Many teams combine both: n8n handles integrations and scheduling, and custom code handles the hard logic behind an API that the workflow calls.

Can n8n run AI agents?

Yes. n8n includes AI agent nodes that connect to models from OpenAI, Anthropic, Google and others, call tools such as APIs and databases, and keep memory between steps. It suits back-office work like email triage, invoice extraction and lead enrichment. For live phone calls, pair a dedicated voice platform with n8n handling the follow-up actions after each call.

Getting it built

Shorai Studios, a founder-led AI studio that builds n8n and custom-code automations, builds in your accounts so you own every workflow from day one. Our business automation service covers n8n builds, CRM automation and integrations. If the workflow needs an AI agent that talks to customers, see our AI agent and voice agent work. For a worked example of an n8n flow with AI extraction and human approval, read our guide to invoice processing automation for small businesses.

Your first version is on us: we build the base workflow for free, so you see it running on your own data before you pay.

Sources

• Zapier pricing
• Jetadmin: Zapier pricing (monthly-billing check, September 2026)
• n8n pricing (official, check for current plans)
• Make pricing (official, check for current plans)
• Jetadmin: n8n pricing overview
• Jetadmin: Make pricing overview