Zapier bills per task (one successful action in a third-party app). Make bills per credit(renamed from “operations” in August 2025, which is now legacy). n8n bills per execution (one complete workflow run regardless of step count). The billing unit, not the monthly fee, is what decides where your cost explodes at scale.
The short version: three ways automation tools count your usage
The cheapest sticker price is often the most expensive to run. A $20/month plan can cost more at scale than a $60/month one, because the unit it meters multiplies with every step you add.
Zapier charges per successful action step. Make charged per operation until August 2025, then shifted to credits, roughly the same mechanic, but AI-heavy modules and iterators now carry variable costs well above the old flat rate. n8n prices by execution: one run equals one charge, regardless of how many nodes fire.
The decoder table below puts all three side by side. If your workflow is already breaking under load, the companion piece on diagnosing automations that keep breaking covers the failure-mode diagnosis. This article decodes the billing units and how to design around them before the cliff arrives.
What counts as a “task” in Zapier (and what is free)
According to Zapier's Help Center (updated June 15, 2026, vendor-authored primary): “A task is any successful action that runs in Zapier. Only successful actions count toward your task usage.”
Each successful action step inside a Zap run costs one task. A 5-action Zap at 10,000 runs per month is 50,000 tasks, not 10,000. Most operators are surprised by this, they assumed the Zap itself was the billing unit.
What does NOT count as a task in Zapier (source: Zapier Help Center, June 15, 2026, vendor-authored primary):
- Trigger steps (Zapier polls apps continuously at no task charge)
- Filter steps and Paths steps
- Formatter, Delay, Looping, Digest, Storage, Zapier Manager, Zapier Tables, Zapier Forms
- Sub-Zap trigger side
- Any action step that errors or halts
- Steps blocked by an upstream filter, path, or error
- Failed search actions set to “No”
- Zapier Agents usage (runs on a separate activity quota)
Zapier's own example: “If a Zap checks an app every two minutes, it will complete over 20 thousand polls per month. Zapier does not charge for these events.”
Multipliers that do bite (source: Zapier Help Center, June 15, 2026; Zapier blog, June 16, 2026, both vendor-authored):
- Zapier Lead Router: 5 tasks per successfully routed lead (not 1)
- Zapier MCP: 2 tasks per successful tool call; failed calls = 0 tasks
- AI by Zapier: Standard = 1x, Advanced = 3x, Premium = 5x
Wire Lead Router to a high-volume trigger without knowing the 5x multiplier and a 10,000-lead month becomes 50,000 tasks before a single downstream action fires.
Make credits (and why “operations” got renamed in 2025)
In August 2025, Make moved from operations to credits. “Operations” still appears as a display metric inside the Make UI but is not the billing meter. Credits are current. Operations are legacy.
The cost change is more targeted than it looks. Michaela from Make's team (Make community, August 2025, vendor-authored): “For customers who do not use Make AI features in their automations, there is no change.” On external API keys: “The dynamic system will only apply to Make's in-house AI modules... if you're asking about an OpenAI module (or an HTTP module that calls the OpenAI API using your own token), nothing changes: 1 credit still equals 1 operation.”
So the pain lands on three specific scenarios: Make's native AI modules, iterators processing large record sets, and paginated API loops. If none of those describe your workflows, the rename is largely cosmetic.
Credit costs per module (Zapier blog, May 12, 2026, vendor-authored, Zapier-biased; re-verify at make.com/en/pricing before publishing):
- Trigger module: 1 credit
- Each action module: 1 credit minimum
- Make Code app: 2 credits per second of execution time
- Make in-house AI modules: variable, well above 1 credit per run
- External AI API key via HTTP module: 1 credit per module run
On router costs:Zapier's blog (May 2026) says filter and router steps each cost 1 credit. Relay.app (June 2026, third-party) calls the router a free built-in exception. The sources conflict; verify on Make's help docs before building high-volume branching scenarios.
The iterator trap is the biggest billing surprise. As reported by Ali Farhat (dev.to, August 2025, third-party): an HTTP call costs 1 credit; an iterator processing 1,000 records costs 1,000 credits. Unused credits expire at the end of your subscription term (Michaela, Make community, August 2025, vendor-authored).
Plan tiers (third-party sources only; make.com/en/pricing not verified directly): Free = 1,000 credits/month; Core = 10,000 credits/month at approximately $9-12/month annual (Zapier blog says $12, Relay.app says $9; verify at make.com/en/pricing).
n8n per-execution model (the outlier that helps at scale)
n8n's pricing page (fetched 2026-06-20, vendor-authored primary): “An execution is a single run of your entire workflow. It doesn't matter how many steps are in the workflow or how much data it processes, it's still a single execution.”
A 30-node pipeline and a 3-node notification both cost one execution per run. Building and testing do not count; only production runs do. Pricing (vendor-authored; prices in EUR): Starter approx. 20 EUR/month / 2,500 executions; Pro approx. 50 EUR/month / 10,000 executions; Community (self-hosted) = free. Zapier's blog (December 2025) cited “$20/month” in USD; verify currency and amount at n8n.io/pricing before citing.
One honest caveat: n8n assumes real technical comfort with JSON and expressions. Per-execution billing also pushes teams toward fewer, larger workflows that can be harder to maintain. Full comparison at n8n vs Zapier for automation workflows. Self-hosting removes metering entirely but adds infrastructure overhead most non-coders should skip.
Zapier
Best for few steps, high volume, predictable billing
- Bills per successful action step
- Triggers and filters are free
- Multi-step Zaps multiply task count
- One charge per whole workflow run
Make
Best for standard integrations, branching scenarios
- Bills per module run (1 credit minimum)
- Own API key stays at 1 credit per run
- Native AI modules cost variable credits
- Iterators charge 1 credit per record
n8n
Best for many steps, moderate volume, technical teams
- One execution per whole workflow run
- Step count never multiplies cost
- Self-hosted community tier is free
- Assumes JSON and expression comfort
The decoder table: tasks vs credits vs executions, side by side
| Platform | Billing unit | What 1 unit counts | What is FREE | What silently multiplies | Free-tier allowance |
|---|---|---|---|---|---|
| Zapier | Task | One successful action step in a third-party app | Triggers (polling); Filters; Paths; Formatter; Delay; Looping; Digest; Tables; Forms; Storage; Zapier Manager; failed/halted steps; blocked steps | Multi-step Zaps (5 steps = 5 tasks/run); Lead Router (5x); MCP (2x); AI steps (Standard 1x, Advanced 3x, Premium 5x) | 100 tasks/month |
| Make | Credit (operations = LEGACY) | Trigger 1 credit; action 1 credit min; Make Code 2 credits/sec; Make in-house AI = variable; external AI API = 1 credit | Error handling modules (Rollback, Break, Ignore); router status disputed (verify) | Iterators (1 credit/record); paginated API loops; Make in-house AI modules | 1,000 credits/month |
| n8n | Execution | One complete workflow run, regardless of step count | Building and testing; all steps per execution; Community self-hosted = completely free | None at step level; total run count only | Cloud: none on free; Starter approx. 20 EUR/month / 2,500 executions |
| Gumloop | Credit (AI-variable) | Base step = 1 credit; standard AI call = 2 credits; advanced AI call = 20 credits; expert AI call = 30+ credits | Logic nodes, filters, loops; standard integrations (Google Sheets, Slack) | AI model tier (advanced = 20x base); triggers cost at least 1 credit (unlike Zapier); credits do not roll over | Approx. 5,000 credits/month (60,000/year) |
Sources: Zapier Help Center (June 15, 2026); Zapier blog (June 16, 2026); Zapier blog on Make pricing (May 2026, Zapier-biased); Make community thread (August 2025); Relay.app blog (June 2026); n8n pricing page (2026-06-20, EUR); Gumloop pricing page + Zapier blog (June 8, 2026). All subject to change.
The table makes the tradeoffs concrete. Few-step workflows at high volume: per-task and per-credit billing are manageable because step count stays low. Many-step workflows at moderate volume: per-execution wins because per-step models compound fast. AI-heavy workflows: the multiplier column matters more than the platform name.
Which billing model is cheapest for your workflow shape
There is no universally cheapest platform. The cheapest one is whichever billing model matches your workflow shape.
| Workflow shape | Best-fit billing model | Why |
|---|---|---|
| Few steps, high volume (1-2 action steps, thousands of runs) | Task (Zapier) or Credit (Make) | Step count stays low so per-step cost does not compound |
| Many steps, low-to-moderate volume (multi-branch, hundreds of runs/month) | Per-execution (n8n) | One execution = one charge regardless of node count |
| AI-heavy (LLM calls inside the workflow) | No credit-per-call model is cheap; watch multipliers on every platform | Advanced AI calls can cost 20x a base step |
The AI-heavy row is where cost surprises land fastest. Gumloop's documented credit costs (Zapier blog, June 8, 2026, vendor-authored): base step = 1 credit, standard model call = 2, advanced = 20, expert = 30+. One advanced call costs 20 times a base step. Make's native AI modules are similarly variable. For AI-heavy workflows, the practical move is to use cheaper model tiers for high-volume simple steps and capable models only where accuracy actually matters. More on matching models to steps at picking the right model for each automation step.
Full platform head-to-head at Zapier and Make compared for agent workflows. Gumloop's credit model in detail at Gumloop vs Zapier: credit-based pricing compared.
Pricing complaints make up half of the Zapier Trustpilot reviews in this sample. The frustration is about opaque billing units, not dollar amounts.
How to design around the pricing cliff (without changing tools)
A few structural changes cut the cost curve without switching platforms.
Filter before you act. In Zapier, Filters and Paths cost nothing. Action steps do. Put filter conditions before action steps so you do not pay tasks on records you will discard. If only 20% of your trigger events need processing, filtering early cuts task count proportionally.
Batch where possible. Processing records one at a time runs one Zap per record. Bundling reduces total runs and total tasks.
Use free Zapier built-ins instead of billed third-party steps. Formatter, Delay, Looping, Digest, Tables, and Storage do not count as tasks. Where they do the same job as a billed third-party action, prefer them.
Audit AI steps first. At 3x-20x the base cost, a single AI step can exceed the combined cost of every other step in the workflow. Downgrade model tier or cache results where accuracy allows.
Know the multipliers before you wire.Lead Router is 5 tasks per lead. MCP is 2 tasks per successful call. Gumloop's advanced model is 20 credits per call. Check these numbers before connecting a high-volume trigger.
For Make: restructure around iterators.Caching stable API responses in Make's Data Store reduces module runs (community suggestion, dev.to, August 2025, not independently verified).
For a fuller picture of total running costs, see how much an AI agent actually costs to run. If you want more sourced breakdowns without vendor spin, subscribe to the newsletter.
Frequently asked questions
What counts as one task in Zapier? A task is any successful action in a third-party app after a trigger fires. Triggers, Filters, Paths, built-in Zapier apps, and failed steps do not count. Source: Zapier Help Center (June 15, 2026, vendor-authored primary).
How are tasks counted in Zapier? Each successful action step is one task. A 5-action Zap uses 5 tasks per run. Multipliers apply: Lead Router = 5 tasks per lead; MCP = 2 tasks per successful call (failed calls are free). Source: Zapier Help Center (June 15, 2026, vendor-authored).
What is a Zapier task vs a Zap? A Zap is the workflow. A task is one successful action step inside it. One Zap run can consume several tasks.
What counts as a credit in Make?Each trigger and action module costs at least 1 credit per run. Make's in-house AI modules cost more (variable by tokens and model). Using your own API key via HTTP module stays at 1 credit per run. Source: Make community thread, Michaela (Make team), August 2025 (vendor-authored).
When using Make's AI Provider, what factors impact dynamic credit usage? Credits vary by tokens consumed, model selected, and API call complexity. Only Make's in-house AI modules use dynamic pricing; external API integrations, including HTTP with your own OpenAI key, stay at 1 credit per run. Source: Michaela, Make community, August 2025 (vendor-authored).
Is Make cheaper than Zapier?Depends on workflow shape. Many-step scenarios at moderate volume can favor Make; high-volume simple flows often favor Zapier's predictable per-task model. Neither is universally cheaper: step count and run frequency decide the real cost.
The honest bottom line
Pick the billing model that matches your workflow shape, not the platform with the lowest monthly fee. Task billing (Zapier): one charge per successful action, predictable for simple high-volume flows, expensive for multi-step ones. Credit billing (Make): one charge per module run, with variable costs on native AI modules, predictable for standard integrations. Execution billing (n8n): one charge per workflow run regardless of step count, structurally cheaper for complex scenarios at moderate frequency, but with a real technical comfort requirement attached.
All prices in this article come from vendor documentation and third-party sources verified as of 2026-06-20. Prices change fast. Check the live pricing page at each vendor before you commit.
If you found this useful, subscribe to the newsletter. Honest sourced breakdowns when there is something worth saying.