Zapier

Section 2 — Is this for you?

Zapier is the right pick when:

• You run a small or mid-sized team
• You need to connect SaaS tools you already use
• Your workflows are simple — a trigger, a few actions, maybe a filter
• You don't have an engineering team on call
• App catalog size matters more to you than workflow complexity

Zapier might work, but watch the cost:

• 10 to 50 people, moderate automation volume
• Multi-step workflows with some branching
• You're OK paying a premium for the biggest app library
• You understand the bill grows with successful actions — overage is charged at 1.25× the plan rate

Look elsewhere when:

• You need EU data residency — Zapier hosts everything in US AWS
• You need self-hosting for compliance, data control, or cost
• You run very high-volume automation and need predictable pricing
• Your workflows need loops over arrays or branching on computed values
• Your engineering team wants code, git, and CI

Section 3 — The real story

Zapier was the first no-code automation platform to reach a mainstream audience. It launched in 2011 out of Y Combinator with a simple pitch: click a button, move data between two apps, no code.

Zapier won by going wide. The platform connects more than 9,000 apps — the largest integration catalog in the automation space.

When a new SaaS product launches, a Zapier integration is usually the first one it ships. If you need to connect a niche tool that came out last month, Zapier is the safest bet.

The cost of that scale

Zapier bills by the task, and tasks are counted as successful action steps. Triggers don't count. Filters, Formatter steps, and Paths don't count either. But every successful action does.

At low volume you won't notice. At tens of thousands of runs a month, the Zapier bill grows faster than most teams expect — and overage is charged at 1.25× the normal task rate.

The workflow model

A Zap is a single trigger followed by a chain of actions. Branching is added through a feature called Paths, available on the Professional tier and up.

Paths support up to 10 branches per group and can nest three levels deep, with a hard ceiling of 100 total steps per Zap. That's flexible, but it's still a tree — not a graph.

Teams that need loops over arrays, or routing based on computed values, often outgrow the model and switch to Make, n8n, or Pipedream.

Where Zapier still wins

Setup. The builder is a browser-based node editor that tests each step live against your connected accounts. You see exactly what data is flowing before you turn the Zap on.

Non-technical users can ship working automations without reading docs. No other platform matches that on-ramp.

The honest summary

Zapier is the default pick for small teams getting started, anyone who needs a long-tail app integration, and workflows where app coverage matters more than pricing predictability.

It's the wrong pick if you need EU data residency, self-hosting, complex loops, or tight cost control at scale.

Section 4 — How it actually works

The facts at a glance

• Compute model: Cloud-only, shared multi-tenant
• Self-host: Not available
• Data residency: US only (AWS). No EU hosting.
• Execution unit: Task — counts only successful action steps
• Builder: Browser-based visual editor with per-step live testing
• Triggers: Instant (webhook) or polling
• Polling interval by plan: Free 15 min · Professional 2 min · Team/Enterprise 1 min
• Branching: Paths — up to 10 per group, nests 3 levels deep, Professional+ only
• Step ceiling: 100 total steps per Zap including all paths

Hosting and data residency

Zapier runs entirely on its own cloud infrastructure in AWS US. There is no self-hosted option and no EU data residency.

GDPR compliance is handled through the EU-US Data Privacy Framework and Standard Contractual Clauses — not through regional hosting. That satisfies the legal requirement, but won't satisfy a buyer whose rules require data to physically sit inside the EU.

How task billing works

The pricing unit is the task. A task equals one successful action step.

Counts as a task:

• Every successful action (send Slack message, create HubSpot contact, etc.)

Does not count as a task:

• Triggers
• Filter steps
• Formatter steps
• Paths steps
• Any step that errors or is skipped

A five-step Zap with one trigger, one filter, and three actions costs three tasks per run.

Go over your plan's monthly allowance and Zapier keeps the Zaps running. Overage is billed at 1.25× the plan's task rate.

The builder

Top-to-bottom node editor in the browser. Each step has a live test runner that calls your connected accounts with real data.

You see exactly what the trigger returned and what the next action is about to send before turning the Zap on.

What's missing: no local development environment, no command-line tool, no native git integration.

Triggers: instant vs polling

Instant triggers use webhooks. The source app pushes new data to Zapier the moment it happens. The Zap fires within seconds, on every plan, including Free.

Polling triggers work the other way. Zapier checks the source app on a schedule. The interval depends on your plan:

• Free: 15 minutes
• Professional: 2 minutes
• Team / Enterprise: 1 minute

Paid plans can also set custom intervals between 1 and 15 minutes.

Whether a trigger is instant or polling is a property of the integration itself — you can't switch modes. Most integrations use polling. Webhooks are the minority.

Branching with Paths

Paths splits a Zap into named branches based on conditions you define. Limits:

• Up to 10 branches per Paths group
• Nests 3 levels deep
• Total Zap capped at 100 steps including every branch
• Available on Professional tier and above

Complex decision trees work. Complex graphs don't.

Section 5 — Where it shines

Zapier is the default pick for small teams getting started with automation. A few things keep it there even as competitors catch up.

The app catalog

More than 9,000 connected apps. Largest in the industry by a wide margin. When a new SaaS product launches, a Zapier integration is usually the first one it ships.

If your stack includes long-tail tools — a niche CRM, a vertical industry app, something that came out last month — Zapier is the safest bet for coverage.

The builder

Browser-based node editor with per-step live testing. You connect an account, pick a trigger, and the editor pulls real data from that account so you can map fields against actual values.

No command line, no local environment, no deploy step. You turn the Zap on and it's running.

Non-technical users ship working automations without reading docs. No other platform matches that on-ramp.

Instant triggers on every plan

Webhook-based triggers fire within seconds, and they're available on every plan including Free. Many competitors gate real-time triggers behind paid tiers.

This matters for any workflow where latency shows up in the user experience — order confirmations, lead routing, incident alerts.

AI features

Zapier rebranded in 2025 as an "AI orchestration platform." Three pieces are worth knowing:

• [Copilot](https://help.zapier.com/hc/en-us/articles/38215656607757-What-is-Zapier-Copilot) — an AI assistant that builds and edits Zaps from a plain-language prompt. Tell it "send a Slack message when a form is submitted" and it drafts the Zap for you
• Zapier Agents — autonomous AI teammates that can call Zap actions as tools and make decisions across steps
• [MCP server](https://zapier.com/mcp) — exposes roughly 40,000 Zapier actions to external LLMs, so tools like Claude can drive Zapier workflows directly

Tables (a simple database) and Interfaces (a form builder) are now bundled into every plan, including Free. That closes a gap Make and n8n used to have over Zapier.

Sub-Zaps for reusable logic

A Sub-Zap is a small workflow you build once and call from any other Zap. Pass data in, get results back. Update the Sub-Zap once and every parent Zap picks up the change.

Sub-Zaps are currently in beta but available to all paid plans. The call itself is free — only the actions inside the Sub-Zap count as tasks.

Autoreplay for reliability

When a step fails, Zapier automatically retries it up to 5 times over roughly 10 hours. Autoreplay is an account-wide setting on Professional plans and up, with per-Zap overrides.

For most transient errors — API timeouts, temporary rate limits — you won't need to do anything. The Zap heals itself.

Section 6 — What breaks most often

Most of Zapier's friction shows up when you scale. What works fine at 1,000 tasks a month starts biting at 100,000.

Task billing gets expensive fast

This is the #1 complaint on review sites. The billing model is simple — one task per successful action — but the bill compounds quickly.

A five-step Zap with three actions, running 10,000 times a month, is 30,000 tasks. A 15-step Zap running 10,000 times could be 100,000+. The Professional plan includes 750 tasks; overage is charged at 1.25× the plan's task rate.

Teams that grow into Zapier often hit a point where Make or n8n offers the same functionality at one-quarter the price.

You can only have one loop per Zap

Zapier's Looping app has real limits:

• 500 iteration ceiling per loop
• One loop per Zap — you can't add loops to multiple Paths branches
• Nested arrays require a custom Code step to flatten first
• Testing only runs the first iteration — bugs show up live

Teams that need to process an array of 2,000 records, or loop inside a branch, usually end up splitting one Zap into several or switching platforms.

Paths is a tree, not a graph

Paths supports up to 10 branches per group and nests 3 levels deep. That covers most decision trees.

What it doesn't do: merge branches back together. Once a Path splits, the branches run independently. If you need a workflow where step 5 in branch A feeds step 6 in branch B, you'll need a second Zap or a different platform.

Rate limit throttling

Zapier enforces its own rate limits on top of the rate limits your connected apps enforce. A few real ceilings:

• Instant triggers cap at 20,000 requests per 5 minutes per user
• Private app API limits: 100 requests/60s on Free and Professional, 5,000 requests/60s on Team and Enterprise
• Webhooks, Code, and Tables each have their own separate limits

Hit the cap and Zapier throttles or returns 429 errors. Throttled tasks can be replayed manually or via Autoreplay, but they won't run on time.

No real staging environment

There's no built-in concept of dev, staging, and production Zaps. The platform doesn't version control your Zaps, doesn't let you promote changes between environments, and doesn't integrate with git.

Teams that need proper change management typically duplicate a Zap, edit the copy, test it by hand, then swap it in. That's fine for occasional changes. It breaks down on a team of 10 people all editing shared Zaps.

Zaps auto-pause when they fail too often

If a Zap errors on 95% or more of its runs over the last 7 days, Zapier pauses it automatically. This is usually a mercy — it stops you from burning tasks on a broken integration.

It's also the kind of thing that will surprise you. If a critical Zap goes down overnight and hits the threshold, it stays off until someone manually turns it back on.

Polling lag on non-webhook integrations

Most Zapier integrations poll for new data rather than receive webhooks. Poll intervals depend on your plan — 15 min Free, 2 min Professional, 1 min Team and Enterprise.

Whether an integration is instant or polling is a property of the integration itself. You can't switch modes. If your source app doesn't offer a webhook, you're stuck with the polling delay.

Section 8 — Getting started reality

Zapier is built for people who have never automated anything before. The path from signup to first working Zap is short on purpose.

The signup flow

• No credit card required for the Free plan
• Email-only signup, then a quick onboarding survey
• You land on a "Create a Zap" screen within about 60 seconds

Your first Zap

The typical sequence:

1. Pick a trigger app and event (e.g., "New form submission in Typeform")
2. Connect your account with OAuth — one click for most apps
3. Pick an action app and event (e.g., "Create row in Google Sheets")
4. Map trigger fields to action fields using the field picker
5. Test each step against real data from your connected accounts
6. Turn the Zap on

Most people ship a working two-step Zap in 15 minutes without reading documentation.

Where people get stuck

• Understanding task billing. New users don't realize every successful action counts toward their monthly cap until they hit it mid-month
• Knowing what to build. The blank canvas problem — Copilot and the template library help, but it still takes practice to see automation opportunities
• Debugging polling lag. When a Zap doesn't fire "instantly," the cause is usually that the source app uses polling, not webhooks

Section 9 — Migration translations

If you're switching from another automation platform, the terminology maps cleanly — but the billing math doesn't always.

From Make

The biggest gotcha: if you're used to Make's operation counting, expect your Zapier bill to scale faster. Zapier tasks aren't cheaper per unit, but they skip the trigger, so the ratio isn't always a clean 30×.

From n8n

Cost delta: n8n self-hosted has no per-action billing at all — if you're coming from self-hosted n8n, expect Zapier's bill to feel steep.

From Power Automate

Power Automate is wired deep into Microsoft 365. If you live in SharePoint, Teams, and Dataverse, you lose native integration depth when you move to Zapier — but gain 8,000+ apps outside the Microsoft stack.

From Pipedream

Pipedream favors developer workflows. Moving to Zapier trades code flexibility for broader app coverage and an easier builder.

Section 10 — AI & MCP readiness

Zapier rebranded in 2025 as an "AI orchestration platform." Three features are worth knowing for buyers evaluating AI capabilities.

Copilot

Status: Generally available

An AI assistant that builds Zaps from plain-language prompts. Tell it what you want and it drafts the trigger, actions, and field mappings. Works across all Zapier products.

Zapier Agents

Status: Generally available

Autonomous AI teammates that can call Zap actions as tools. Agents reason about a task, decide which actions to run, and execute them across the 9,000+ connected apps.

MCP server

Status: Beta

The Zapier MCP server exposes roughly 40,000 Zapier actions to external LLMs via the Model Context Protocol. Tools like Claude or ChatGPT can trigger Zapier workflows directly.

Available to all accounts, but still in beta — early reports call it powerful but inconsistent. Not production-ready for critical workflows yet.

The summary

Zapier's AI story is ahead of Make and Pipedream and competitive with Microsoft's Power Platform. If AI-driven automation is a key buying criterion, Zapier is a strong pick. If you need a fully stable MCP integration today, wait a quarter or pick a different tool.

Section 11 — Compliance at a glance

The facts

• SOC 2 Type II: Yes — audited annually by an AICPA-authorized auditor
• SOC 3: Yes — public summary report available
• HIPAA: No — Zapier does not sign BAAs and does not support processing PHI
• GDPR: Yes — handled via EU-US Data Privacy Framework and Standard Contractual Clauses
• Data residency: US only (AWS) — no EU hosting option
• Trains AI on your data: No — customer data is not used to train Zapier's AI models
• [SSO (SAML 2.0)](https://help.zapier.com/hc/en-us/articles/8496279747085-Set-up-single-sign-on-with-SAML): Included on Team and Enterprise plans
• SCIM user provisioning: Enterprise only
• Custom data retention: Enterprise only

The fine print

Zapier's SOC 2 Type II status covers its security controls and is reviewed every year. The Trust Center publishes the updated report after each audit cycle.

HIPAA is the hard "no." Zapier explicitly does not support workflows that touch Protected Health Information. Healthcare teams can still use Zapier for non-PHI workflows (marketing, HR, billing that doesn't include patient data), but anything involving patient records needs a different platform.

GDPR compliance is contractual, not architectural. Data still lives in US AWS. If your compliance rules require data to physically sit in the EU, Zapier won't meet them regardless of the DPF and SCCs in place.

For AI features, Zapier commits in its privacy policy that customer data is not used to train its own AI models. Third-party AI integrations (OpenAI, Claude, etc.) are defaulted out of their training pipelines, though individual users can opt in.