What is a Next Gen Zap?

A Next Gen Zap is a workflow that an AI tool builds for you and Zapier runs. You describe the work in plain language to your AI tool, such as Claude, ChatGPT, or Cursor, or to Zapier in the Workflow Manager. Your AI tool or Zapier writes the workflow, and Zapier runs it after you close the chat.

Next Gen Zaps run alongside your Classic Zaps. Your Classic Zaps keep running unchanged, and you choose when to import one. A Next Gen Zap is also different from a Code by Zapier step in a Classic Zap. A Code by Zapier step is a single step, and a Next Gen Zap is a complete workflow that Zapier runs from start to finish.

By the end of this article, you'll understand how a Next Gen Zap is built and run, when to use one, and how it compares to a Classic Zap.

Available on plans:

Free

Professional

Team

Enterprise

Early Access

Next Gen Zaps is in early access by request. Sign up at https://zapier.com/workflow/early-access. If you’re on a Team plan, the request must come from an account owner or admin. If you’re on an Enterprise plan, contact Support for access. This feature has limited availability and is in active development.

Where to find it

You can find Next Gen Zaps in the Zapier left sidebar, under Automations. This page lists your Next Gen Zaps. Click a workflow to open it in the Workflow Manager.

How Next Gen Zaps work

You build a Next Gen Zap, publish it, then Zapier runs it. After each run, Zapier keeps a record of what happened, and an agent can help when a run fails.

Build

To build a Next Gen Zap, describe what the workflow should do. Your AI tool or Zapier chooses the app connections, writes the steps, and tests the workflow with you. You can build one in Claude, ChatGPT, Cursor, or another external AI tool by connecting it to Zapier through Zapier MCP. To build without Zapier MCP, describe the workflow to Zapier in the Workflow Manager.

The workflow is written in code, but the Workflow Manager displays it as a flow diagram and has a TypeScript code editor, so you can review each step or edit the code.

You can also import a Classic Zap that you own or can edit. The import is a copy, and the original Classic Zap keeps running until you turn it off, so both can run and do the same work twice.

Publish

Your changes are saved as a draft. When you publish the draft, it becomes a new version of the workflow, and Zapier runs that version. Learn more about Next Gen Zap drafts and versions.

Run

A Next Gen Zap runs on Zapier, not inside your AI tool. It starts when something happens in your apps or on a schedule, without a chat window open. The event that starts it is its trigger. When the workflow is published and turned on, you can also run it on demand. Ask your AI tool, click Run in the Workflow Manager, or call it from your own code with the Zapier SDK. It runs the same way no matter which AI tool you used to build it.

Tip

The trigger can also be a webhook, so another app or system can start the workflow by sending data to it. A webhook receives events only while the workflow is turned on.

During a run, the workflow uses your app connections in Zapier. It can also send custom API requests through those connections, so you do not need to add separate credentials.

The workflow uses AI only in the steps that need judgment, such as an AI by Zapier step. The other steps run the same way every time.

Because the logic is code, a workflow can:

  • Loop through a list without the 500-item limit of Classic Zaps.
  • Run loops inside loops.
  • Split into branches that rejoin later.
  • Catch an error in one step and continue the run.
  • Pause between steps, or until another system sends a callback. Each wait can last up to 1 year.

Maintain

Every Next Gen Zap appears in Zapier, whichever AI tool built it. Zapier records each run, step by step, in the workflow's run history.

If a run fails, such as after a field is renamed, an agent can diagnose the problem and prepare a fix. This is called agentic management, and it is on by default. If a run fails because a connection expired, the agent tells you which connection to reconnect. You choose whether the agent applies fixes on its own or asks you first. Even when it applies fixes on its own, it asks for your approval when a fix is higher risk or relies on a guess about how your workflow should work. When the agent applies a fix, it publishes the change as a new version of the workflow. When optimization is on, the agent reviews successful runs of agent steps and suggests how to make them more predictable. It does not apply these suggestions on its own. Learn more about Agentic management in Next Gen Zaps.

When to use a Next Gen Zap

If you ask your AI tool to do the same task every week, or copy its results into other apps, a Next Gen Zap can do that work for you. It is also a good fit when a workflow needs to do things like process a whole list, nest loops, rejoin branches, handle errors, or wait a long time.

The following ideas show a few of the things you can build.

What to try Example prompt What it shows
Process a whole list "Find every HubSpot contact untouched for 90+ days and add them to a re-engagement list." No 500-item loop cap. For very large lists, process in batches instead of all at once so you do not trip the connected app's rate limits. Start with a few hundred records and scale up.
Reconcile two systems "Check every row in my Sheet against my Stripe charges and flag mismatches." Loop your full dataset and compare as you go.
Sync that survives failure "On a closed Salesforce deal, create records in QuickBooks, Asana, and Slack. Skip any that fail and report back." Catch an error and keep going.
Route, then merge "Send enterprise leads to a rep, the rest to nurture, then save all to HubSpot and post to #leads." Branches split and rejoin.
Score and route leads "Score each new lead, send high-priority leads to the right rep, and add every lead to Salesforce." An AI step scores each lead, then branches route it.
Triage support tickets "When a new Zendesk ticket arrives, pull the customer's history, suggest a reply, and assign the ticket to the right person." An AI step suggests the reply. The other steps run the same way every time.
Loops inside loops "For each Airtable client, total their open QuickBooks invoices into one digest." Nested loops with a summary at the end.
Long follow-up that can stop "Email trial users on day 3, 7, and 14. Stop if they upgrade." Waits of up to a year between steps.

A Next Gen Zap is not the right fit for:

  • A one-off task. Ask your AI tool to run the action through Zapier MCP instead.
  • Work that must respond in under a second.
  • Code that must run inside your own network.

Example

Example

You want every new lead from your website form scored, routed by priority, and saved to Salesforce.

  • You tell your AI tool: "Score each new lead, send high-priority leads to the right rep, and add every lead to Salesforce."
  • Your AI tool builds a Next Gen Zap that starts when someone submits your form. An AI step scores the lead from the details in the form.
  • The workflow branches on the score. A high-priority lead is assigned to a rep, and every other lead is enrolled in a nurture sequence. Both branches then add the lead to Salesforce.
  • A few weeks later, Salesforce renames a field, and a run fails. Agentic management diagnoses the failure and prepares a fix. With Automatically fix selected, it publishes the fix as a new version. It asks for your approval first if the fix is higher risk or relies on a guess about how the workflow should work. You can review the change in the version history.

Only the scoring step uses AI. The other steps run the same way every time, and the run history records how each lead was scored and routed.

Next Gen Zaps and Classic Zaps

Both kinds of Zaps connect your apps and run on Zapier. They differ in how you build them and how their logic works.

Feature Classic Zap Next Gen Zap
How you build it You set up a trigger and actions manually in the Zap editor and map fields between steps. You describe the workflow, and your AI tool or Zapier writes the Zap for you. You can also edit the code.
What it is made of A trigger and one or more action steps. Code that Zapier runs from start to finish. It usually starts with a trigger, followed by steps and logic.
How logic works You add steps such as Paths, Filters, and Looping by Zapier. A loop can run up to 500 times. The logic is written in code. Loops have no 500-item limit, and branches can split and rejoin.
When a run fails Zapier records the error in your Zap history, where you can troubleshoot it and replay the run. Zapier records the error in the run history. Agentic management can diagnose it and prepare a fix, or you can troubleshoot it with your AI tool.

Learn more about Classic Zaps.

Security considerations

Zapier keeps the credentials for your app connections. Your AI tool uses those connections through Zapier and does not receive the credentials.

Use judgment about which Zapier account, app connections, and data you make available while testing Next Gen Zaps. If an agent has access to private data, is exposed to untrusted instructions, and is allowed to send data elsewhere, it could take actions you did not intend, including sharing sensitive information.

Start with low-risk workflows and app connections. Avoid connecting accounts that contain highly sensitive data, such as legal documents, financial records, or private company files.

If you want a stricter safety boundary, use a separate Zapier account or less-sensitive app connections. If your Enterprise account uses workspaces, you can also use a dedicated test workspace.

Limitations

Next Gen Zaps have limits on code size, how long a run can wait, sharing, and how long Zapier keeps your data. Read Next Gen Zap limitations for the full list.

Next steps

You now know how a Next Gen Zap is built, published, and run, and how it compares to a Classic Zap. Now:

Was this article helpful?
0 out of 0 found this helpful