The Zapier app exposes the whole catalog as a single action. You choose an endpoint from a dropdown, Zapier redraws the input fields to match it, and the step runs whenever your trigger fires.
Connecting
Add an action step, search for APIVerve, and sign in when prompted. The connection resolves to your account's primary API key — you do not paste it into a step, and it is not visible in the Zap afterwards.
Two consequences worth knowing up front:
Sub-keys will not work here. A sub-key sent to a connector is
rejected with a 401 and Sub-keys cannot be used for integrations. Connections use the
primary key.
Rotating your key breaks the connection. Every saved Zap using it starts failing at the same moment. Reconnect once in Zapier and every Zap picks the new connection up.
Once made, the connection is reusable across every Zap in the account — you connect once, not per Zap.
Choosing an endpoint
The dropdown lists the catalog alphabetically by title, grouped by category. It is generated from the live catalog, so an endpoint that shipped this morning is already there.
File-upload endpoints — image conversion, OCR, QR code reading and the rest — are filtered out, because a Zapier step cannot send multipart form data. That is 9 of 367. For those, use Zapier's Webhooks by Zapier action against the endpoint directly.
Filling in fields
Once you pick an endpoint, Zapier asks for that endpoint's parameters and nothing else. The fields are built from the same schema the API validates against, which means:
- Required fields are marked required and Zapier will not let you publish without them.
- Parameters with a fixed set of values become dropdowns rather than free text, so you cannot typo an enum.
- Example values arrive as defaults, so a step is usually runnable before you have mapped anything — useful for testing.
- A read-only info panel sits at the top with the endpoint's description, its category, what it costs per call, and a link to its reference page.
Map fields from earlier steps as you would anywhere in Zapier. A parameter that takes a list or an object is a text field — pass JSON.
Reading the output
Output fields are generated from the endpoint's sample response, so they are available for mapping before the step has ever run.
Nested values are flattened with a double underscore. A response shaped like
{ "data": { "location": { "city": "Austin" } } } gives you data__location__city. Arrays get
[] in the path. There is also always a Raw Data (JSON) field holding the whole data
object as a string, for when you want to pass the response somewhere else intact rather than
picking it apart.
Every step also returns status and error, which is what to branch on — see below.
What it costs
Browsing endpoints, changing your selection and inspecting output fields are all free. Only running the step costs a call, at that endpoint's rate — the same as calling it directly.
Two habits save real money:
Put a Zapier filter before the APIVerve step, not after. A filter that runs after the call has already spent the call. Filtering first is the difference between validating every row in a spreadsheet and validating the ones that changed.
Watch what a schedule multiplies into. A Zap on a 15-minute schedule is about 2,880 calls a month before it has done anything interesting. Turn on the 80% usage alert in settings.
What to build
Validate before it reaches your CRM. New HubSpot or Salesforce contact → Email Validator → write the result back to the record, and branch invalid ones into a review list. This is the most common shape and the one with the clearest payback, because the alternative is discovering the bad addresses at send time.
Enrich a form submission. Form submitted → IP Lookup for location, or Company Lookup on the email domain → append to the row before it reaches your sales queue.
Generate an asset on creation. New product row → QR Code Generator → save the image to Dropbox and write the link back.
Screen a signup. New signup → Disposable Email Checker and Username Profanity → route anything suspicious to a manual queue. There is a full walkthrough of this one in block fake signups.
When it fails
Errors come back with the API's own status code and message, unchanged. So the error handling page applies here exactly as it does to a direct call, and the status is the fastest way to tell what went wrong.
401 — the connection is broken, not the step. Either the key was rotated, or you are trying
to use a sub-key. Reconnect the account.
400 — a parameter is wrong, and the message names it. Usually a mapped field arriving empty
because the trigger did not supply it, or arriving in the wrong shape. Zapier shows the request
it sent in the step's data tab; compare it against the endpoint's reference page.
429 — either you are over the per-minute rate limit or the account is out
of calls. A Zap replaying a backlog is the usual cause, since Zapier will run through queued
tasks as fast as it can. Check the balance in analytics first; if there
is balance left, it is the rate limit and Zapier's autoreplay will clear it.
The step times out. Connector calls are cut off at 25 seconds. A handful of endpoints — screenshots, PDF rendering, anything that drives a browser — can exceed that on a heavy page. For those, call the endpoint from a Webhooks by Zapier step, where you control the timeout.
The step ran but sent no parameters. For endpoints that accept POST, the connector sends a POST only when at least one parameter has a value; with everything empty it falls back to a GET. If a step behaves as though your mapping was ignored, check whether the mapped field was actually populated on that run.
Next
Make covers the same ground with branching, Pipedream is the one to reach for when a step needs real code, and all endpoints is the catalog the dropdown is built from. Integrations compares all of them side by side.