Zapier

Updated October 2026. Prices checked 28 October 2026.

Zapier's free plan covers unlimited Zaps, Tables and Forms, but caps you at 100 tasks a month and two-step Zaps, so serious use starts on a paid plan and the bill follows your task volume. That single fact decides most buying choices. The rest of the pitch, an AI orchestration layer that dates back to a 2011 launch, is where the platform has changed most.

How it works in practice

The core mechanic is unchanged. A trigger in one app (a form submission, a new Slack message, a row added to a sheet) fires actions in others. You build in the visual editor, in code, or by describing the job to Zapier Copilot, which maps fields and suggests steps. The newer route is to start in Claude, ChatGPT or Cursor through Zapier MCP and have Zapier turn the result into a saved workflow.

That last route is the Next Gen Zaps pitch, still in early access. Zapier calls its approach Fluid Determinism, and it describes four levels: fully deterministic steps, deterministic steps with one AI task, a deterministic flow with an embedded agent, and a fully agentic setup where an agent plans across tools and asks you to review. The idea is that repeat AI work gets converted into fixed steps that use fewer tokens, and AI is added back only when something unexpected happens. It's a sensible design. You pay for model calls only where judgement is needed.

On the free plan, Zaps poll every 15 minutes, so a new lead can sit unseen for a quarter of an hour. Paid plans add multi-step Zaps, webhooks, unlimited premium apps, AI fields and conditional form logic.

What stands out

MCP as a governed action layer. Zapier says one connection reaches Claude, ChatGPT, Cursor and 230+ other AI tools, with access to 9,000+ apps and 66,000+ triggers and actions. You don't hand the AI tool passwords or API keys. Permissions are set per team member, so an assistant that drafts a CRM update can't also touch invoicing unless you allow it. Building one custom integration per AI client is the alternative, and for most teams that's a worse deal.

Self-healing Zaps. Zapier says every Zap gets an AI manager agent that detects failures and fixes them. That claim is the most ambitious on the site and it's marked early access, so treat it as a roadmap item. The practical benefit if it works: fewer silent failures in long workflows, which is the usual complaint with any automation tool once nobody owns the run history. Early-access Next Gen Zaps are available on paid plans and don't count toward your tasks for now.

IT-side controls. The governance page lists action restrictions, managed connections, AI guardrails, asset history, log streaming and a bring-your-own-model option. Policies apply at app, action and connection level, and IT can switch a workflow off in one click. Compliance covers SOC 2 Type II, SOC 3, GDPR and CCPA, with SAML SSO and SCIM for provisioning. Zapier's own customer stories give the scale it targets: Mercari US reports 47,000 tickets a month resolved by Zapier, and Okta reports 13% of Workforce Identity escalations handled automatically. Those are vendor-published figures, so read them as upper-end examples, not a baseline.

Where it falls short

The free plan's limits bite quickly: two-step Zaps only, one seat, 100 tasks and 15-minute polling. Live chat support doesn't start until the Professional tier at 2,000 tasks, so a small team on the entry tier is on email support.

Every step in a Zap and every external connector call uses tasks, and Zapier says rates vary by AI model tier, code runtime and connector type. A heavier model or a longer code step costs more tasks than a plain action. Predicting a bill before you've built the workflow is harder than the headline price suggests.

Connector depth varies. 9,000+ apps is the count, but some connectors expose a full API and others only a few triggers and actions. Check the exact trigger and action you need for your two or three key apps before paying.

Next Gen Zaps will move to the standard task-based pricing at general availability, and Zapier says any pricing structure changes will be shared before billing starts. That leaves the self-healing and cost-saving claims unproven at scale and the final price unknown. Agents and chatbots also sit under Add-ons on the pricing page, separate from the platform plans.

Value for money

As of October 2026, Professional starts at $19.99 a month and Team at $69 a month, both with yearly billing, which Zapier says saves 33% against paying monthly. Those are entry prices for the lowest task tier. Task volume is the lever that moves the bill.

Here's a worked case at 10,000 tasks a month. A three-step Zap uses 3 tasks per run, so 10,000 tasks is about 3,333 runs. A ten-step workflow with a lookup, a formatter and an AI step uses 10 or more per run, so the same allowance covers 1,000 runs, roughly 33 a day. Zapier's pricing slider has a 10K setting, and you should read the monthly figure there for your plan before committing, because that number, not $19.99, is what you'll pay. All of it draws from one shared pool across Zaps, AI steps, code, MCP and the SDK, so an agent that polls often can use up the budget your core workflows need.

For a five-person team the seat count doesn't change the price on Professional, and Team stays at $69 for up to 25 users with shared Zaps, folders and app connections plus SAML SSO. So Team earns its price on collaboration and security, not on volume. If your workload is high-volume with heavy branching, price the same job in Make, which bills by operation, before you commit.

Who should pick it, who should skip it

Pick Zapier if you're a non-technical operations, support or revenue team that needs many different apps connected quickly, or an IT group that wants AI assistants to act on business systems under one permission and audit layer. The 9,000+ app library and the MCP connection to 230+ AI tools are the reasons. The free plan is enough to prove one two-step automation before you spend anything.

Skip it if your workflows run thousands of steps a day, need intricate branching, or must run on your own infrastructure. Developers who want code-level control will do better with n8n, and teams weighing licences against hosting costs can compare the options in Open Source Zapier Alternatives by Licence and Self-Hosting Cost. Also skip it for now if you're buying specifically for self-healing Zaps, since that feature is in early access.

Setup notes

Tasks are one shared pool, so check the task usage rates page before you build an agent or code step, then watch the usage figure for the first week. If you want to try Next Gen Zaps, you need a paid plan, and builds during early access don't count toward your task total.