Skip to content
ConsultEvo

Zapier Pricing in 2026: Compare Tasks, Make Credits, and n8n Executions

There is no universally cheapest automation platform. Zapier may suit teams that prioritize hosted, no-code automation and broad app coverage. Make may suit teams that want visual scenario control and credit-based usage. n8n may suit teams that can operate hosting and prefer execution-based pricing. The fair comparison is the same workload, mapped to each platform’s billing rules and operating responsibilities.

This guide was checked on November 17, 2026. Pricing, product availability, currencies, plan features, and billing behavior can change. For example, a workflow handling 10,000 monthly leads might search for an existing CRM record, classify each lead with AI, and write accepted records. Those steps are not automatically 10,000 equivalent billing units across the three platforms.

Use the article as a decision and estimation method, not as an account-specific quote or a promise that a particular integration will support every required action.

Which automation platform is the better value for your workload?

Start with the operating model rather than the headline price. Choose Zapier as a candidate when hosted infrastructure, non-developer ownership, app coverage, governance, and quick deployment matter. Consider Make when the team wants visual scenario control and can estimate module-level credit use. Consider n8n when execution-based billing fits the workflow and the organization can operate or contract for hosting.

Then test the actual workload. Count source events, reads, successful writes, branches, AI calls, tool calls, retries, data volume, and human review. Add support, governance, hosting, monitoring, and maintenance. A platform is better value only relative to those requirements and the people responsible for running it.

Compare the same workload and operating responsibilities, not the billing unit names or headline prices.

What Zapier pricing includes, and what to verify

Zapier’s pricing page currently displays Free at $0, Professional starting at $19.99 per month, Team starting at $69 per month, and custom pricing for Enterprise. These are displayed starting prices, not universal quotes. The selected task tier, billing frequency, currency, and account configuration affect the actual price. Do not assume that Professional always includes a fixed 750 task allowance. Check the task tier shown for the account.

  • Free: 100 tasks per month and two-step Zaps, with access to products including Tables and Forms.
  • Professional: Multi-step Zaps, premium apps, webhooks, AI fields, conditional form logic, and additional workflow features. The task allowance and support terms depend on the selected tier.
  • Team: Collaboration features such as shared Zaps and folders, shared app connections, SAML SSO, and priority support, subject to the current plan terms.
  • Enterprise: Advanced administration, app controls, observability, annual task limits, and plan-specific support and governance terms, with custom pricing.

Zapier says pay-per-task billing can allow workflows to continue after the included allowance if that setting is enabled. Otherwise, workflows pause until the next usage period. Confirm the selected allowance, whether pay-per-task billing is enabled, the overage rate, and the pause behavior before buying. Check the live plan comparison for the support and governance controls required by your organization.

What counts as usage: tasks, credits, and executions

A Zapier task is generally a successful action in a Zap. Trigger steps do not count, and failed or halted action steps generally do not count. In the documented Zap task model, Filter by Zapier, Paths by Zapier, Formatter by Zapier, Delay, Looping, Sub-Zap, Digest, Storage, Tables, and Forms are listed as not consuming normal task usage. Product-generation differences apply, so check the rules for the specific Zap or product. AI, MCP, SDK, and external calls may have separate billing treatment. See Zapier’s task measurement documentation.

Make credits are consumed by module actions. Reads, searches, creates, updates, transformations, aggregations, and iterations may use credits, while documented exceptions apply to some routers and error handlers. Some AI features have variable usage. Make’s pricing page explains the broad model, but the exact estimate depends on the modules and features in the scenario.

n8n hosted executions count runs of an entire workflow rather than individual steps. A short run and a complex run can each count as one execution, but their runtime, data volume, concurrency, and infrastructure demands can differ.

Platform Usage unit Measure in the workflow
Zapier Successful actions, with product-specific exceptions Which actions succeed, and which AI or external calls have separate treatment?
Make Credits for module actions, with exceptions Which reads, writes, transformations, and iterations run?
n8n hosted Whole-workflow executions How many runs, retries, replays, and concurrent workloads occur?

Keep a workload ledger with separate counts for source events, reads, successful writes, AI calls, tool calls, retries, review cases, and workflow runs. Translate each count into the relevant vendor’s rules. A task, credit, and execution are not interchangeable.

How to build a workload-normalized comparison

Use this hypothetical monthly baseline: 10,000 inbound events, one lookup per event, one AI classification per event, a CRM write for accepted records, a 5% retry rate, a 2% human-review rate, and three AI tool calls for cases requiring enrichment. The volume is illustrative, not a vendor benchmark. A 5% retry rate would represent 500 retry attempts if every retry occurred once, but the actual count depends on the failure and replay design.

For a practical lead workflow, use a website form as the source, the CRM as the system of record, and a named sales operations or CRM owner for exceptions. A representative sequence is: receive the form event, validate required fields, look up a potential CRM match, classify unstructured details if needed, validate the result, and update or create an accepted record. Route conflicting matches, missing fields, and uncertain classifications to human review.

The following is a proposed implementation model, not a vendor-published template. Confirm the exact app actions, fields, permissions, and rate limits before designing around them.

Trigger AI job Validation Action and fallback
Website form submission with a stable source event ID Classify unstructured lead details and optionally summarize enrichment Require an approved classification, confidence from 0 to 1, required CRM fields, and an unchanged source ID Upsert or update the CRM record when deterministic rules pass. Send missing fields, conflicting matches, and low-confidence results to the sales operations queue.
Support ticket supplied to an MCP-compatible client Search connected helpdesk, issue-tracker, and knowledge-base context; suggest category and next step Confirm ticket and workspace identity and require source references for recommendations Show the result to a support representative. Route cases without relevant documentation to review and do not send a customer-facing response without a separate approval rule.
Lead captured from a form, webinar, or website Summarize enrichment and suggest a routing category Normalize email and company fields, preserve original values, and check CRM-native identifiers Create or update the CRM record only after deterministic matching. Route conflicting company matches or missing required data to review.

A compact illustrative classification result could look like this:

{
  "classification": "qualified_lead",
  "confidence": 0.82,
  "reason_code": "service_fit",
  "source_event_id": "evt_illustrative_001",
  "review_required": true
}

The ID and values are illustrative. Validate that the classification belongs to an approved list and that confidence is numeric and between 0 and 1. Keep the AI suggestion separate from authoritative CRM fields. Deterministic rules should control deduplication, permissions, required fields, and irreversible routing.

01Build the workload ledgerCount events, reads, writes, branches, AI calls, tool calls, retries, replays, and review cases. The workflow owner records the assumptions and expected monthly volume.
02Map usage by platformApply Zapier task rules, Make credit rules, and n8n execution rules independently. Record uncertain app-specific costs for verification rather than guessing.
03Add operating costsInclude hosting, monitoring, backups, security work, support, review time, maintenance, and the labor needed to investigate failures.
04Get owner sign-offHave the system owner approve the exception path, data responsibilities, deduplication method, and estimated volumes before committing to a plan.

For Zapier, a successful lookup or write may affect task usage under the documented action model. For Make, map each scenario module to its credit behavior. For n8n, count whole-workflow runs, including how retries and replays are implemented. A self-hosted n8n estimate must also include infrastructure, monitoring, backups, security, scaling, and engineering time. The hypothetical baseline does not contain enough account-specific information to produce a reliable cross-platform monthly price.

How AI changes the usage estimate

AI availability does not mean unlimited AI usage. Zapier documents AI by Zapier model rates of Standard 1x, Advanced 3x, and Premium 5x. Tool calls add usage at the selected model rate. AI by Zapier is available on Professional, Team, and Enterprise, while Free restrictions prevent publishing a Zap containing an AI by Zapier step. Check the current AI model-tier and task rules for the selected plan and configuration. MCP and third-party AI app calls have their own product and billing considerations.

Decision point

In a hypothetical Zapier run, one Premium-tier AI model call plus three tool calls uses (1 + 3) × 5 = 20 tasks under the documented formula. Record the model tier and tool-call count in the ledger, then recheck current product and plan rules before purchase.

Use deterministic filters for required fields, duplicate IDs, account ownership, high-risk categories, suppression lists, and exact status transitions. Reserve AI for interpreting unstructured text, summarizing a transcript, locating relevant knowledge, or suggesting a classification. Parse and validate the output before a CRM write. Send malformed or uncertain results to a person rather than allowing the model to decide permissions or irreversible actions.

Choose a platform by operating model, not headline price

  • Zapier: A candidate when hosted infrastructure, non-developer ownership, app coverage, governance, and deployment speed matter. Zapier currently states that its platform connects to more than 9,000 apps. Confirm the exact action, fields, permissions, and rate limits required because an app count does not prove that a particular workflow is supported.
  • Make: A candidate when visual scenario design and granular module control fit the team’s way of working. Estimate credits for the actual reads, writes, transformations, iterations, and other module actions. Teams evaluating that operating model can review Make automation services.
  • n8n: A candidate when execution-based pricing and unlimited workflow steps suit the workload and the organization can operate or contract for hosting. The current hosted Starter plan is listed at €20 per month billed annually for 2,500 executions, and Pro at €50 per month billed annually for 10,000 executions. Check the current n8n pricing page for availability and terms. Self-hosting transfers infrastructure and operational responsibilities to the customer or service provider.

Use three questions to narrow the choice: Who will maintain the system? Which billing unit maps most clearly to the workload? What infrastructure, support, security, and governance responsibilities is the organization prepared to own? Teams assessing a hosted no-code approach can also review Zapier automation services.

Customer results: useful patterns, not price benchmarks

Published customer stories can show the shape of a workflow, but their savings are reported results, not independent benchmarks. Zapier reports that ClickUp reduced support-ticket research from about 15 to 4 minutes across roughly 5,000 monthly tickets, which the story calculates as 917+ hours saved monthly. Remote reports 616 hours saved monthly and $500,000 in avoided IT hiring costs. These figures do not predict another organization’s return without its own baseline, labor cost, volume, implementation cost, quality measures, and review effort.

The ClickUp story describes an MCP-based support-triage pattern involving ticket context, issue-tracker and knowledge-base searches, AI classification, and a structured result for a support representative. The Remote story describes multi-channel help-desk intake, identity context from Okta, AI triage, Notion ticket creation, Zapier Tables, Zapier Agents, and Slack assignment. Neither page is a complete, ready-to-import template. Review the ClickUp story and Remote story as pattern evidence only.

For ROI, measure baseline and post-launch handling time, error and rework rates, review volume, exception resolution time, and maintenance effort. Subtract implementation, subscription, review, and operating costs from the value of the measured change.

Data integrity controls that affect real cost

Retries can create duplicate CRM records if two runs both search, find no match, and then create. A lookup followed by create is not race-safe under concurrent runs. Use a stable source event or record ID and, where the destination supports it, a database-enforced unique key, destination-native idempotency key, or transactional upsert. Serialize writes by entity when necessary.

For a lead workflow, a proposed unique key might be source_system + source_lead_id + enrichment_version. For a support workflow, it might be source_system + source_record_id + workflow_version + action_type. These are illustrative data-model recommendations, not built-in Zapier fields. If multiple runs can occur for the same record, include a run identifier or action version at the declared row grain rather than relying on date alone.

Preserve original lead data separately from enriched and AI-generated values. Retain the source record identifier, source URL or document ID, retrieval time, provider or app, model tier, prompt version, validation result, reviewer decision, and destination record ID where those fields are relevant. Track accepted records, prevented duplicates, review rates, failed runs, replays, and time to resolve exceptions.

A pre-purchase checklist for a reliable cost estimate

Confirm before rollout
  • Verify the plan tier, task or credit allowance, billing frequency, overage setting, and pause behavior.
  • Test the exact app actions, permissions, fields, rate limits, and data volume required by a representative workflow.
  • Record the AI model tier, tool-call count, required output fields, prompt version, and human-review threshold.
  • Define the system of record and stable source identifier. Use destination-enforced uniqueness or transactional upsert when concurrent runs are possible.
  • Include support, governance, review labor, and, for self-hosted n8n, infrastructure and maintenance ownership in the total-cost estimate.

Do not select a platform until one representative workload has been mapped to its billing units, retry and exception behavior, ownership, data controls, and total operating cost. That estimate, rather than a comparison of isolated starting prices, is the useful basis for a purchase decision.