Skip to content
Plugins and integrations5 min read

Zapier and Make

Point a Call action at a Zapier Catch Hook or a Make webhook, and the chatbot can start your automations mid-conversation.

Browse topics

What it is

Zapier and Make are automation tools: they sit between apps and move information around when something happens. Agentency connects to them the same way almost everything connects to them — with a webhook. You create a webhook in Zapier or Make, copy the address it gives you, and add it to your chatbot as a Call action. When a conversation needs that tool, the chatbot calls the address and your Zap or scenario runs.

This is not a website plugin. Nothing in this article puts a chat bubble on your site — that is the Plugins gallery. Call actions run in the other direction: the chatbot reaching out to your systems, mid-conversation.

When you would use it

Typical jobs people wire up this way:

  • Create a row in a CRM or a spreadsheet when someone asks to be contacted.
  • Post to a Slack or Teams channel when a conversation looks like a complaint.
  • Create a ticket in a help desk the chatbot cannot answer for.
  • Look up a booking or an order in a tool that has no public API of its own, using the automation platform as the glue.

If all you want is the leads themselves, you may not need any of this: leads captured in chat already appear under Customers, and you can export them. Reach for an automation only when something has to happen now.

Where to find it

Open Dashboard → Chatbots → your chatbot → Actions.

Before you start

  • Your plan must include at least one Call action. Free includes none at all. See Which plans include integrations and the pricing page for what each tier includes.
  • You need an account on Zapier or Make with permission to create a webhook trigger.
  • Decide what information the chatbot should send. Keep it to what the automation actually needs.

Steps

  1. In Zapier, create a Zap with the Webhooks by Zapier → Catch Hook trigger and copy the URL it shows you. In Make, create a scenario starting with the Custom webhook module and copy its address.
  2. In Agentency, open Dashboard → Chatbots → your chatbot → Actions.
  3. Create an HTTP Call action and paste the address into it. See Create a Call Action for the full field-by-field walkthrough.
  4. Describe, in plain language, when the chatbot should use it. This description is what the chatbot reads to decide — "use this when a customer asks to speak to sales and has given their email" works far better than "sales webhook".
  5. Define the pieces of information it should send, with a clear name for each.
  6. Save and enable the action.
  7. Open Test Chatbot and say something that should trigger it.
  8. Check Call action history on the same tab to see whether the call fired and what came back, then confirm the Zap or scenario ran on the other side.

What you will see

Call actions is a table of the tools this chatbot may use, each with a name, an address, and an on/off switch. History on the same tab shows each attempt with its result, which is the first place to look when someone says "the automation did not run".

In Zapier or Make, your webhook trigger shows the request as soon as the chatbot calls it, so you can build the rest of the Zap against real data rather than a guess.

Limits and plan notes

Each chatbot has a limit on how many Call actions it can hold, and that limit rises with your plan. Free has none. Check the pricing page for the current numbers, or open Billing to see what your plan allows.

A few rules worth knowing before you build:

  • The chatbot decides when to call. There is no "run this on every conversation" switch. If it never fires, the description is usually the reason.
  • The address must be a normal, public HTTPS address. Requests to internal or private network addresses are refused on purpose, so an automation you host yourself must be reachable from the internet.
  • Secrets belong on the action, not in your instructions. Anything you write into instructions is text the model sees; credentials go in the action's own configuration.
  • Zapier and Make have their own quotas. A busy chatbot can consume Zapier tasks or Make operations quickly. Filter early in the Zap rather than letting every call count.

MCP connectors — a different way of giving a chatbot tools — may or may not be switched on for your account. See MCP connector availability.

Common problems

I cannot create the action.

You are at your plan's Call action limit, or you are on Free, which has none. Turn off an action you are not using, or upgrade.

The Zap never fires.

Work through it in this order: is the action enabled; does Call action history show an attempt; if it shows an attempt, did the call succeed; if it shows nothing, rewrite the description so it clearly states the situation that should trigger it.

It fires too often.

The description is too broad. Narrow it to the specific condition, and add an instruction telling the chatbot to collect the details it needs before calling.

The Zap runs but the fields are empty.

The names of the pieces of information the action sends must match what your Zap expects. Re-run the Zapier trigger test after the chatbot has called it once, so Zapier learns the real shape.

I want two-way — Zapier starting a conversation.

That is the other direction, and it needs a key rather than a webhook. See Custom API and your own backend.

Common questions

Is this the same as the website Plugins tab?

No. Plugins put a chat bubble on your site. Zapier and Make are the other direction: the chatbot calling out to your tools.

Can I start a Zap for every new lead?

There is no run-on-every-conversation switch. The chatbot calls the action when the conversation needs it, or you export Customers and process the file.

My Zap never fires. Where do I look?

Call action history first. No attempt means the description is too vague; a failed attempt means the address or the payload is wrong.

Can I use a self-hosted automation server?

Yes, if it is reachable on a normal public HTTPS address. Requests to internal or private network addresses are refused on purpose.

Was this article helpful?

Ready to try it on your own content?

Create a free workspace, add a document, and ask the questions your team is tired of answering.