API

The API block makes an HTTP request to a URL and returns the response. Use it to call any REST API: fetch data, create a record, or trigger an external endpoint.

API

URL-

MethodGET

Query Params-

Headers-

Body-

error

Configuration

URL

The endpoint to call. Type a static URL, or insert a connection tag to build it from an earlier output, like https://api.example.com/users/<start.userId>.

Method

The HTTP method: GET, POST, PUT, DELETE, or PATCH. Defaults to GET.

Query Params

Key-value pairs appended to the URL as a query string. apiKey and limit become ?apiKey=…&limit=10.

Key-value request headers, such as Content-Type: application/json or Authorization: Bearer <secret>. Standard headers like User-Agent and Accept are added automatically, and your values override them.

Body

The request payload for POST, PUT, and PATCH, sent as JSON. Type it directly, or pull it from an earlier output with a connection tag.

Advanced

Timeout (ms). How long to wait before giving up. Defaults to 300000 (5 minutes), up to 600000 (10 minutes).
Retries. Number of retry attempts on timeouts, 429 responses, and 5xx errors. Defaults to 0.
Retry delay / Max retry delay (ms). The exponential-backoff bounds used between retries.
Retry non-idempotent methods. Off by default, so POST and PATCH are not retried, which avoids duplicate writes. Turn it on only when a repeated request is safe.

Outputs

After the request completes, later blocks read its result by name:

Output	What it is
`<api.data>`	The response body, parsed as an object for JSON or returned as text otherwise
`<api.status>`	The HTTP status code, like `200` or `404`
`<api.headers>`	The response headers, as an object

Branch on the result by reading <api.status> in a Condition block.

Example

A workflow that calls an HTTP endpoint and summarizes the response:

The API block builds its URL from the Start input, fetches the data, and the Agent reads the response as <api.data>.

Best Practices

Keep secrets in environment variables. Reference them with {{VAR}} in the URL or headers; never hardcode keys.
Handle failures. Read <api.status> and branch on it with a Condition, or connect the error path for network failures and timeouts.
Set retries for flaky endpoints. Use the Advanced retry settings for idempotent calls. Leave POST and PATCH off unless a repeated request is safe.
Reference only the field you need. Pull <api.data.id> rather than the whole <api.data> when the response is large.

Common Questions

The default timeout is 300,000 milliseconds (5 minutes). You can configure it up to a maximum of 600,000 milliseconds (10 minutes) in the block's Advanced settings.

Retries are attempted for network and connection failures, timeouts, rate-limit responses (HTTP 429), and server errors (5xx). Client errors like 400 or 404 are not retried.

Retries use exponential backoff starting from the configured retry delay (default 500ms). Each subsequent retry doubles the delay, up to the maximum retry delay (default 30,000ms).

No. POST and PATCH are non-idempotent, so retries are disabled for them by default to avoid creating duplicate resources. You can enable retries with the 'Retry non-idempotent methods' toggle in Advanced settings, but be aware this may cause duplicate requests.

Standard headers such as User-Agent, Accept, and Cache-Control are added automatically. Any custom headers you configure are merged with these defaults, and your values override automatic headers with the same name.

The block sends JSON request bodies through the UI. The underlying HTTP tool also supports form data: if you pass form-data parameters, it constructs a multipart/form-data request automatically. For most cases the JSON body field is sufficient.

Common Questions

On this page