Zapier

What is Zapier?

How does Zapier work?

Use Cases

Pricing

Pricing model: Freemium

Common Limitations

• Per-task billing scales linearly with step count, making multi-step workflows increasingly expensive · Zapier bills per task, and every successful action step consumes one. A five-step workflow (trigger plus four actions) running 1,000 times a day uses 4,000 tasks per day, or 120,000 tasks in a 30-day month. On Zapier's annual Professional plan, the 200,000-task tier costs $769 per month; the same workflow on Make's annual Core plan costs $99, and on n8n Cloud Pro €120—or nothing on self-hosted n8n Community Edition.[reference:0]The structural problem is not the headline price but the meter: the more steps a workflow legitimately needs, the more it costs, even if each step is trivial. Zapier's own architecture encourages splitting logic across paths and filters to reduce billable steps, but branching is exactly where simple two-step automations break down and where users move to paid plans. At 10,000 tasks the Professional plan costs $129/month annually; at 100,000 tasks $489; at 1,000,000 tasks $2,199.[reference:1]A ten-step workflow running 10,000 times a month is 90,000 billable tasks—roughly $450/month on Zapier before any AI steps. The pricing model is aligned with simple, low-step, low-volume workflows maintained by non-technical users. It is poorly aligned with the multi-step, high-volume pipelines that teams eventually build.
• Overage is automatic, penalized, and then hard-paused at 3x allocation · Exceeding a plan's task limit does not simply stop the account. Zapier switches automatically to pay-per-task billing at 1.25x the base rate on annual plans and 2.5x on monthly plans.[reference:2]This continues until usage reaches 3x the subscribed task allocation, at which point further Zap runs are held—paused, not lost, but unavailable until the next billing cycle or an upgrade.[reference:3]The sequence matters: a spike in usage produces a bill at a penalized rate first, then a service interruption. For teams that treat Zapier as infrastructure, this is a double failure mode—cost overrun followed by downtime. The 1.25x/2.5x multipliers and the 3x cap are documented, but they are not surfaced prominently during plan selection, and the effective cost of a mid-cycle spike is often discovered on the invoice rather than in advance. Turning off idle Zaps, choosing the task tier from real historical usage, and paying annually are the only practical mitigations, and all three require ongoing discipline.
• The free tier cannot prototype a realistic workflow · The Free plan's 100 tasks are generous only if the workflow is two steps. Multi-step Zaps—any workflow with more than one action—require Professional. Filters and Paths, which are essential for any conditional logic, are locked behind paid plans.[reference:4]This means the free tier cannot answer the question most evaluators actually need answered: “will this multi-step workflow work?” A user can test a trigger-action pair, but cannot test branching, multi-app chains, or the task consumption of a realistic process. The 15-minute polling interval further limits time-sensitive prototyping.[reference:5]Zapier offers a 14-day Professional trial with no credit card, which is the practical evaluation path, but the free tier itself is a demo of the editor, not of the product's core capability. Teams that start on Free and scale into multi-step automation should budget for the Professional tier from the outset rather than treating Free as a viable long-term home for anything beyond a two-step personal Zap.
• AI steps multiply task consumption and can pause the Zap at 75 tasks per run · Since June 15, 2026, AI by Zapier steps are priced by model tier: Standard 1x, Advanced 3x, Premium 5x, and bring-your-own-model 1x. The default is Advanced (3x), so a new AI step immediately triples its task cost per run before any tool calls.[reference:6]Tool calls add to that base cost at the same rate. A single AI step that reaches 75 tasks in one execution pauses the entire Zap and waits for manual approval—a circuit breaker against runaway agentic cost, but also an operational interruption for workflows that are supposed to run unattended.[reference:7]The 75-task threshold is per execution, not per billing cycle, and it applies even if the account has ample task allowance remaining. Teams using AI steps at volume should expect periodic pauses and design a notification and approval path around them. Adding one AI step at the Advanced tier to a workflow running 30,000 times a month adds 90,000 tasks, pushing a 200,000-task plan to the 300,000-task tier at $1,069 per month.[reference:8]The AI step's cost is not a marginal add-on; it can change the entire plan tier.
• No EU data residency, and SSO/SCIM are gated behind Team and Enterprise · Zapier stores customer data on AWS servers in the United States (US-East 1) and does not offer EU-only data residency on any plan.[reference:9]It certifies to the EU-US, UK, and Swiss Data Privacy Frameworks with standard contractual clauses as a fallback, so this is a residency limitation rather than a compliance gap—but for organizations with data-sovereignty requirements, it is a hard boundary that no plan tier removes. Make lets organizations pick a US or EU data zone; n8n's hosted plans store data in Frankfurt and can be self-hosted anywhere.[reference:10]Separately, SAML SSO begins at Team ($69/month annual), and SCIM provisioning is Enterprise-only.[reference:11]A single-department Professional account cannot enforce SSO or centralized user lifecycle management. The governance features that regulated or security-conscious organizations require are structurally separated from the tier that most teams start on.
• Zap size caps and the structural mismatch between task billing and execution-based competitors · A single Zap is capped at 100 steps total and can nest at most 3 Path steps, each with up to 10 branches.[reference:12][reference:13]For most workflows this is not binding, but for long, branching pipelines it forces splitting logic across multiple Zaps—each of which may consume its own tasks and trigger its own polling. The deeper issue is the billing unit itself. Zapier bills per action step; Make bills per credit (per-module operations); n8n bills per execution (one full workflow run regardless of step count).[reference:14]For a five-step workflow running 1,000 times a day, Zapier costs $769/month on annual Professional, Make costs $99 on annual Core, and n8n costs €120 on Cloud Pro—or nothing self-hosted.[reference:15]At low volume the difference is noise; at scale it is the entire cost decision. Zapier's advantage is breadth (9,000+ apps) and ease of use, but that advantage carries a structural cost premium that grows with workflow complexity. The premium is justified for simple, low-step automations maintained by non-technical users. It is increasingly difficult to justify as step count and run volume rise.
• Documented security research on MCP introduces a new attack surface · Zapier's MCP server extends its 9,000+ app connections to external AI clients, which is a genuine capability advantage. It also expands the attack surface. A June 2026 fwd:cloudsec presentation titled “Zapocalypse” demonstrated an exploit chain that escalated to a full Zapier AI platform account takeover, exposing hardcoded MCP keys and high-privilege tokens, with a stored XSS via dependency poisoning of a core design system affecting authenticated sessions of over 8,000 third-party integrations.[reference:16]Zapier's MCP security documentation states that MCP is enabled by default for all accounts, including Enterprise, and that app and action restrictions cannot currently be set exclusively for MCP—any restrictions apply account-wide.[reference:17][reference:18]Research on agent loops has also flagged that MCP tools which create records can be invoked in loops by poorly designed agents, creating thousands of duplicate records.[reference:19]These are not reasons to avoid Zapier's MCP, but they are reasons to treat it as a governed integration surface rather than a convenience toggle: scope permissions tightly, enable only the tools the agent needs, and monitor for runaway loops rather than assuming the default configuration is safe.

Zapier vs. Make and n8n

This comparison uses official pricing pages, documentation, and independent cost analyses available as of November 2026. It focuses on the structural difference that determines the invoice eighteen months after adoption: what each platform counts as a billable unit. Zapier bills per task (every action step), Make bills per credit (per-module operations, renamed from operations), and n8n bills per execution (one full workflow run, regardless of step count). Capability has converged; the meter has not. The choice is less about features and more about which billing model matches the workflow's shape and volume.

Comparison checked against official product information on 2026-11-04. Features and pricing may change.