n8n vs Make vs Zapier for AI Agents: What Each Really Costs and Which to Pick

n8n vs Make vs Zapier for AI agents

You build a lead follow-up workflow in Zapier over a weekend. It catches the web form, looks up the contact, has an AI write a reply, sends a text, logs the lead, pings you in Slack and updates your CRM. It works. Then you turn on ads, it runs a thousand times a month, and the plan you need jumps from $29.99 to $193.50 a month. Moved to n8n, the same workflow fits a $24 plan.

That gap is not about features. It comes from how each tool counts usage, and once you understand that, the choice between n8n, Make and Zapier for an AI agent gets much simpler. This guide explains the billing, prices one real workflow on all three, and gives you a way to decide this week.

The one difference that matters: what each tool counts

n8n counts executions. An execution is one run of a whole workflow, no matter how many steps it has. A 3-step workflow and a 30-step workflow cost the same per run. That is why n8n gets cheaper the more your workflow does, and why AI agents, which loop and call tools, fit it well.

Make counts credits, which replaced operations in August 2025 at one credit per operation. Every time a module performs an action, it uses a credit, and that includes the trigger reading from a webhook or app. Routers and error handlers are free. If a step loops over ten items, that step uses ten credits.

Zapier counts tasks. Each successful action step is a task. Triggers, filters, formatters, paths and delays are free, and failed steps do not count. AI steps use a variable number of tasks depending on the model. So a Zap with six real actions costs six tasks per run.

What the same workflow costs on each

Take the lead follow-up workflow above: one trigger plus six steps, run 1,000 times a month. The table shows the result, using each vendor's published prices as of October 2026.

On n8n it uses 1,000 executions, which fits the Starter plan at 2,500 a month: $24 billed monthly, or $20 a month billed annually. On Make it uses about 7,000 credits (seven modules times 1,000), which fits the Core plan at 10,000 credits: $16 billed monthly or $12 annually. Make is the cheapest of the three here, as long as nothing in the scenario loops. On Zapier it uses 6,000 tasks, which pushes you past the 5,000 tier to the Professional 10,000 tier: $193.50 billed monthly or $129 annually.

Now double the workflow to 14 steps and nothing changes on n8n. Make doubles to about 14,000 credits and needs a bigger tier. Zapier doubles to 12,000 tasks. Steps are where Zapier and Make get expensive, and AI agents add steps.

None of these include the AI model itself. On n8n and on Make with your own API key, you pay OpenAI or Anthropic directly for tokens, which for short replies is usually pennies a day. Make's built-in AI provider charges credits for tokens on top of the operation credit. Zapier's AI steps cost extra tasks.

Building an actual AI agent in each

n8n has a dedicated AI Agent node. It works as a tools agent: you connect a chat model (Anthropic, OpenAI, Google Gemini, Groq, OpenRouter, or a local model through Ollama) and at least one tool, such as a calendar, a database lookup or another workflow. The agent decides which tool to call. This is the most flexible of the three and the one most builders on this site use.

Make has its own AI Agents, and since November 2025 every paid plan, including Core, can connect your own AI provider to them. With your own provider, each agent operation costs one credit and you pay the provider for tokens. If you already think in Make's visual scenarios, it is a reasonable choice for simple agents.

Zapier's agent product, Zapier Agents, is billed separately in activities rather than tasks: 400 activities a month free, or 1,500 a month on the Pro plan at $33.33 a month billed annually, as of October 2026. Zapier also offers an MCP server on every plan that lets an AI assistant like Claude call your Zapier actions, at two tasks per tool call. Zapier's strength is its catalog of app integrations, not its agent depth.

What builders on this site actually use

Of the 117 case studies on this site, 29 use n8n, 19 use Make and 3 use Zapier. That skew is real and it tracks the billing. One builder replaced $3,000 a month of manual work with 12 n8n workflows covering lead management, onboarding and newsletters, and swapped a $150 a month uptime monitor for a free n8n workflow (builtwithagents.ai/strategy/12-n8n-workflows-replaced-3000-month-manual-work). Another runs n8n with local models on a single server at $0 a month in software fees (builtwithagents.ai/strategy/self-hosted-ai-automation-n8n-docker-local-llms-zero-cost).

Make shows up in lighter builds. A freelancer sells $497 voice agents built on Make's free plan with Claude's free tier (builtwithagents.ai/strategy/voice-ai-agent-497-free-tools-local-businesses), and several of the voice receptionists here use Make to connect Vapi or Retell to a calendar. Zapier appears mainly as glue, as in a moving company's lead response agent that used Zapier to pass leads to the AI (builtwithagents.ai/strategy/openclaw-lead-response-agent-blue-collar-businesses).

Which one to pick

Pick Zapier if you are not technical, your workflows have two to five steps, your volume is low, and you need an app that only Zapier connects to. You pay more per run in exchange for the easiest setup and the biggest catalog. Below a few hundred runs a month, the difference may be $10 to $20, which is not worth an extra weekend of learning.

Pick Make if you like building visually, your scenarios are short and do not loop over many items, and you want the cheapest hosted plan for moderate volume. Watch iterators: a step that processes 50 rows uses 50 credits.

Pick n8n if you are building a real AI agent, your workflows have many steps or branches, your volume is growing, or you want to self-host. It has the steepest learning curve of the three, but it is the only one where cost stays flat as the workflow grows.

One warning for agencies. n8n's free self-hosted Community Edition is under the Sustainable Use License, which allows internal business use. It does not allow you to host n8n and charge clients for access to it, white-label it, or let people outside your organization build workflows in your instance without a commercial license. Building and consulting on n8n for clients is allowed. The common pattern is to set up each client in their own n8n account or instance. Read n8n's license FAQ before you build a business model on it.

How to choose this week

Step 1: Write your workflow on paper

List the trigger and every step, including lookups, AI calls, messages and logging. Estimate how many times a month it will run, and note anything that loops, like "for each line item" or "for each new row."

Step 2: Price it on all three

Multiply it out using the rules above. n8n is one execution per run. Make is one credit per module per item. Zapier is one task per action step. Then find the smallest plan that fits each, with 50 percent headroom for growth.

Step 3: Check your three must-have apps

Search each tool's integration directory for the three apps you cannot live without, such as your CRM, your calendar and your phone or SMS provider. If one tool lacks a native integration, you will need HTTP requests and API docs, which is fine for a builder and a wall for a non-technical owner.

Step 4: Build the smallest version

Build the trigger, the AI step and one action on your top choice. Make's free plan gives 1,000 credits a month and Zapier's free plan allows two-step Zaps only. For n8n, either pay for a month of Starter at $24 or self-host the Community Edition on a small server, such as DigitalOcean's 2 GiB droplet at $12 a month as of October 2026.

Step 5: Run it for a week and read the usage page

Push real or test data through it for a week, then open the usage or billing page. Compare actual executions, credits or tasks against your paper estimate. If the real number is double, find the loop that caused it.

Step 6: Add error handling before you rely on it

Every tool fails sometimes: an API times out, the AI returns something odd, a field is empty. Set up an error workflow or error route that alerts you by email or Slack. An AI agent that fails silently is worse than no agent.

What it costs

As of October 2026: n8n Cloud Starter is $24 a month billed monthly ($20 annually) for 2,500 executions, and Pro is $60 a month ($50 annually) for 10,000. The self-hosted Community Edition is free for internal use, and you pay for the server. Make Core is $16 a month ($12 annually) for 10,000 credits, Pro is $28 ($21), and the free plan gives 1,000 credits. Zapier Professional starts at $29.99 a month ($19.99 annually) for 750 tasks, and its 10,000-task tier is $193.50 ($129). Add your AI model costs and any SMS, voice or data tools on top.

Where each one breaks

Zapier breaks on cost as workflows get longer, and on complex logic, where Paths and Filters get messy. Make breaks quietly when credits run out: scenarios stop until the next cycle, so set a usage alert. n8n Cloud Starter caps a single execution at five minutes, which can bite a slow multi-step agent, and self-hosting makes you responsible for updates, backups and uptime. If nobody on your team will own a server, use n8n Cloud.

Frequently Asked Questions