Every endpoint in the catalog can be driven from an automation platform without writing a request by hand. Pick an endpoint from a dropdown, map fields from an earlier step, and the platform makes the call.
There are two ways that happens, and the difference matters more than the platform you pick.
Built connectors versus calling the API
Four platforms have a connector we build and maintain, backed by a dedicated endpoint on our side: Zapier, Make, n8n and Power Automate. These give you a searchable list of endpoints and input fields that change to match whichever one you selected. Nothing is hand-maintained — the list is generated from the same catalog the API runs on, so a new endpoint appears in the dropdown when it ships, not when we get round to updating a connector.
Everywhere else you call the REST API directly. Pipedream, ViaSocket and LangChain each have a page here because the idiom differs, but the underlying request is the one on any reference page. Any platform with an HTTP step can do this, so the absence of a page is not the absence of support.
What every built connector does
The four share one implementation, so they behave the same way in the places that count.
| Step | What it does | Credits |
|---|---|---|
| Connect | Validates your key, returns your plan and email for the connection label | Free |
| Choose an endpoint | Lists the catalog, alphabetically, with categories | Free |
| Fill in fields | Reads the endpoint's schema and builds inputs from it | Free |
| Map the output | Reads a sample response and builds the output field list | Free |
| Run | Calls the endpoint | 1 call |
Only the run costs anything. Browsing the catalog inside Zapier, changing your mind about which endpoint you want, and inspecting output fields are all free — which makes the connector a reasonable place to explore, not just to execute.
File-upload endpoints are excluded from every connector dropdown, because none of these platforms send multipart form data through a generic step. That is 9 of the 367 in the catalog. Everything else is there. To use one of the excluded endpoints from an automation, call it with an HTTP step instead — see making requests.
What an integration call costs
Exactly what the same call costs anywhere else. A connector run re-issues your request to the public API with your own key, so it bills one call at that endpoint's rate, counts against the same rate limit, and returns the same body — including the same error with the same status code when something is wrong.
There is no separate integration plan, tier or quota. Every plan can use every connector:
| Plan | Credits / month | Rate limit |
|---|---|---|
| Free | 100 | 5/min |
| Starter | 100,000 | 60/min |
| Pro | 500,000 | 180/min |
| Mega | 2,000,000 | No limit |
This is worth stating plainly because it is easy to assume otherwise: a workflow that runs every five minutes is 8,640 calls a month whether you built it in Zapier or in cron. The platform does not change the arithmetic. Set a schedule you can afford and turn on the usage alert in settings.
Which key each platform takes
It depends on whether the platform runs an account-connection step.
Zapier, Make and Pabbly need the primary key. Those three link an account before any step
runs, and that connection refuses a sub-key with a 401:
{ "status": "error", "error": "Sub-keys cannot be used for integrations" }n8n and Power Automate accept a sub-key. They store a key and send it on each call, with no separate linking step, so the credential is validated the same way a direct API call is — and a scoped sub-key is the better thing to put in a shared n8n instance or a Power Platform environment.
Where you must use the primary key, bound what the automation can reach in the workflow rather than in the key.
The practical consequence is that your primary key ends up stored in a third-party platform. That is fine, and it is what every connector on every platform does, but it makes rotation a real event: rotating the key breaks every saved connection until you update it. Know how many platforms hold it before you rotate.
Choosing one
Zapier if you want the shortest path and the widest app library, and you are happy on a hosted platform.
Make if your workflow has real branching in it. The visual builder handles a multi-step scenario better than a linear Zap does, and its module inputs arrive pre-filled with the endpoint's example values.
n8n if you want to self-host, or if you want more than API calls — the n8n node also reaches JSON bins and your usage analytics, which the other three do not.
Power Automate if you are inside Microsoft 365 and need Teams, SharePoint or Dataverse in the same flow.
Pipedream if you would rather write the eight lines of Node yourself and keep full control of retries and error handling.
Next
Each page below covers connecting, the fields you will see, and the failure modes specific to that platform. If you are not using a platform at all, making requests is the direct route.