n8n vs Zapier vs Make Pricing: Count the Right Unit

n8n, Zapier, and Make price different units. Zapier counts successful task-producing steps. Make counts credits, with most module actions using one credit. n8n Cloud prices monthly workflow executions with unlimited steps, while n8n also offers a self-hosted Community Edition. These product facts were checked on August 12, 2026 against the official Zapier task rates, Make pricing page, and n8n pricing page.

There is no honest winner without a workflow. Draw one run, count each platform’s billable unit, add AI provider charges, and price the labor needed to build and operate it. Self-hosted n8n can remove a cloud subscription from the calculation, but it does not remove hosting, backups, upgrades, security, or support work.

What does Zapier bill?

Zapier states that a successful standard step uses one task. Several built-in tools use zero tasks. Its current rate page also lists different multipliers for AI tiers and programmatic access. For example, Zapier MCP uses two tasks per tool call, while standard, advanced, and premium AI steps use one, three, and five tasks respectively. These are current product rules, not permanent assumptions, so check the official Zapier task rates before buying.

Count the whole path. A trigger may start a workflow, but later actions create the billable work. Branches, loops, AI tool calls, and retries can make two apparently similar automations consume different task totals.

Do not estimate from the number of incoming records alone. Run a sample and export usage. Divide total tasks by accepted business outputs. A task that completes successfully can still produce an output the business rejects.

What does Make bill?

Make uses credits. Its pricing page says each module action typically consumes a credit, although some advanced AI features can use more. The page listed a free plan with up to 1,000 monthly credits and paid plan examples based on 10,000 monthly credits when checked on August 12, 2026. See Make’s official pricing and credit explanation for current figures.

A scenario can consume several credits for one business result. Reading a record, transforming data, calling another app, and updating a destination can each be separate actions. Iteration can multiply that count.

Map the scenario with a normal record, an exception, and a retry. Then use Make’s own execution data to calculate credits per accepted output. Plan prices can change, and annual billing can differ from monthly billing, so store the date and billing term beside any quote.

What does n8n bill?

n8n Cloud prices workflow executions rather than every step. Its official pricing page describes executions with unlimited steps and lists a standard self-hosted Community Edition on GitHub. Paid self-hosted Business and Enterprise options have separate features and licensing. Check the current n8n plans rather than assuming every self-hosted deployment has the same rights or support.

Execution-based billing can favor long workflows because adding another node does not automatically create another billed execution. Workflow design still matters. Sub-workflows, polling schedules, retries, concurrency, saved execution data, and retention can affect capacity and operations.

Community Edition software is not the same as a managed service. The business operates the server and database. It also owns upgrades, backups, monitoring, credentials, incident response, and recovery. The self-hosted n8n cost guide provides a worksheet for those costs.

How do the billing units compare?

This comparison uses vendor descriptions checked on August 12, 2026. It does not compare quoted monthly totals because volume, billing term, region, tax, and plan tier can change the result.

The units are not interchangeable. Ten thousand credits, tasks, and executions can represent very different amounts of completed work.

How should you price one workflow?

Choose a business output. Suppose the workflow receives an order, checks inventory, creates an invoice, sends an internal notice, and updates the customer record. This is an example workflow, not a platform benchmark.

For each platform:

1. Draw the normal path and every branch.
2. Mark actions that create a billed unit.
3. Add loops and average items per loop.
4. Add observed retries and failed runs.
5. Add model tokens or calls charged by another provider.
6. Divide monthly cost by accepted orders.

Use a test account or vendor estimator where available. Do not infer charges from a marketing diagram. Save screenshots or exports with the date for procurement records, but do not rely on them forever.

The AI automation ROI calculator can compare the resulting cost per accepted output with measured labor and review savings.

How do AI steps change the comparison?

AI adds at least two cost layers: the automation platform’s unit and the model provider’s unit. A platform may also bundle selected model usage or apply a different unit multiplier to an AI step.

Record input tokens, output tokens, tool calls, and retries during the pilot. If the agent chooses tools at runtime, use a distribution rather than one happy-path count. A request that calls three tools is not the same as a request that calls none.

Review time belongs in the same calculation. A cheap model step that creates more corrections may cost more per accepted output. The small-business LLM API budget guide shows how to set a variable usage budget without guessing a universal model price.

When is self-hosting n8n cheaper?

Self-hosting can make sense when workflow volume is high, the team already operates reliable infrastructure, or control over deployment is worth the work. It can also cost more when one person becomes an informal on-call operator.

Price:

• Compute, database, storage, backups, and network traffic.
• Setup, upgrades, testing, and recovery drills.
• Monitoring and incident response.
• Security patches and credential management.
• Time spent supporting workflow authors and users.
• Paid n8n features or support, if required.

Use a loaded hourly cost for internal labor. Mark server prices as quoted on a specific date. Do not call staff time free because it does not appear on a card statement.

Which platform fits a small business?

Zapier may fit a team that wants a managed product and can express the workflow with supported app actions. Make may fit a team that wants visual control over branching and can track credits closely. n8n Cloud may fit a technical team that prefers execution-based pricing. Self-hosted n8n may fit a team prepared to operate the whole service.

Those are fit tests, not performance claims. Integration quality, permissions, regional needs, support, and staff familiarity can outweigh unit pricing.

Run a bounded pilot on the same workflow in the serious candidates. Compare accepted outputs, staff time, failure recovery, and total cost. A platform that is cheaper per unit can still be more expensive if it takes longer to maintain.

What should you check before choosing?

Check the current official pricing page on the day of purchase. Confirm the billing term, included volume, overages, data retention, execution history, support, and export options.

Then ask operational questions:

• Who receives a failure alert?
• Who can change credentials and workflows?
• How will a bad deployment be rolled back?
• Can usage stop at a hard limit?
• How will the business export workflows and logs?
• Which sensitive actions still need human approval?

The final decision should show cost per accepted output under normal and stress volumes. If the comparison ends at the plan’s monthly sticker price, it has not counted the right unit.