n8n HTTP Request node: a practical reference

The n8n HTTP Request node calls any REST API from a workflow. It is the node you reach for when n8n has no dedicated app node, when an app node lacks the operation you need, or when you are replacing axios and fetch calls from another platform. This page covers every setting that matters in practice, with configurations you can copy, and the fixes for the errors people hit most often.

If you are moving from Pipedream, the mapping from Send any HTTP Request and axios is a separate page. This one is the n8n reference.

What the node does

One HTTP Request node sends one request per input item. If the previous node outputs 50 items, the node sends 50 requests and outputs 50 results. That is the most common surprise for people coming from code: you don't write a loop, the item list is the loop.

The response body becomes the output item's json. When an API returns a top-level JSON array you usually get one item per element; when the list is nested (for example under data), add a Split Out node to turn it into separate items.

Node parameters

Parameter	Options	Notes
Method	DELETE, GET, HEAD, OPTIONS, PATCH, POST, PUT
URL	Any URL, expressions allowed	https://api.example.com/users/{{ $json.id }}
Authentication	None, Predefined Credential Type, Generic Credential Type	n8n recommends predefined when it exists
Send Query Parameters	Using Fields Below, Using JSON
Send Headers	Using Fields Below, Using JSON
Send Body	Form URLencoded, Form-Data, JSON, n8n Binary File, Raw	Content type picked per body

All values accept expressions, so {{ $json.email }} or {{ $('Webhook').first().json.body.orderId }} work in the URL, headers and body.

Authentication

The docs list two families:

• Predefined Credential Type. Reuses a credential n8n already supports (Slack, GitHub, HubSpot and hundreds more). This is how you call an endpoint that the app node doesn't cover: same credential, custom request.
• Generic Credential Type. Basic auth, Custom auth, Digest auth, Header auth, OAuth1 API, OAuth2 API, Query auth and Simplified Custom Auth.

For a typical API key, choose Header Auth with name Authorization and value Bearer sk_live_..., or whatever the API documents. The key is stored encrypted as an n8n credential and doesn't show up in the workflow JSON when you export or share it. A header typed into Send Headers in plain text does.

Body types

Body Content Type	Sends	Use for
JSON	application/json	Most REST APIs
Form URLencoded	application/x-www-form-urlencoded	OAuth token endpoints, legacy forms
Form-Data	multipart/form-data	File uploads with extra fields
n8n Binary File	The file as the whole body	Raw uploads (S3 presigned PUT, for example)
Raw	Any content type you set	XML, plain text, GraphQL as text

With JSON > Using JSON, wrap the whole object in an expression when it contains dynamic values:

{{
  {
    "email": $json.email,
    "tags": ["migrated", "pipedream"],
    "amount": $json.amount
  }
}}

The common issues page recommends exactly this pattern for the "JSON parameter need to be an valid JSON" error.

Import cURL

Most API docs ship cURL examples. In the node's Parameters tab select Import cURL, paste the command and select Import. n8n fills in method, URL, headers and body. Two caveats from the docs:

1. It overwrites the existing node configuration.
2. Parameter values are imported as strings. If the API needs numbers or booleans, switch the body to Using JSON and paste the object.

After the import, move any Authorization header into a credential. Otherwise the key stays in the node and travels with every export.

Node options

Select Add Option to reach these.

Option	What it does
Array Format in Query Parameters	foo=bar&foo=qux, foo[]=bar or foo[0]=bar
Batching	Items per Batch and Batch Interval (ms)
Ignore SSL Issues	Downloads the response even if certificate validation fails
Lowercase Headers	On by default
Redirects	Follow redirects (on by default), Max Redirects
Response	Include Response Headers and Status, Never Error, Response Format
Pagination	Off, Update a Parameter in Each Request, Response Contains Next URL
Proxy	Overrides HTTP_PROXY / HTTPS_PROXY / ALL_PROXY
Timeout	Milliseconds to wait for the response headers

Response handling

By default the node returns only the body and fails on any non-2xx status. Two switches change that:

• Include Response Headers and Status returns headers, statusCode and body.
• Never Error returns success for every status code.

Turn both on when the workflow must branch on the status itself, for example "create if 404, update if 200". Put an If node after it that checks {{ $json.statusCode }}.

Response Format is Autodetect by default. Choose File for PDFs and images (the file lands in the binary field you name in Put Output in Field), or Text to keep the raw string.

Pagination

Pagination is the reason many people keep a code step around. n8n handles the common cases without code. The pagination cookbook gives three patterns.

Page number in the query

1. Pagination Mode: Update a Parameter in Each Request
2. Type: Query, Name: page
3. Value (expression): {{ $pageCount + 1 }}

$pageCount starts at 0, so the first request asks for page 1. Set the page size in the main Send Query Parameters section (for example limit=100), not inside the pagination settings.

Next URL in the response

1. Pagination Mode: Response Contains Next URL
2. Next URL: {{ $response.body["next-page"] }}

Adjust the path to whatever your API returns: $response.body.links.next, $response.body.paging.next.link, and so on.

Page in the body

For POST-based search APIs: Method POST, Pagination Mode Update a Parameter in Each Request, Type Body, Name page, Value {{ $pageCount + 1 }}.

Variable	Meaning
$pageCount	Pages fetched so far, starts at 0
$request	The request the node sent
$response	The response: $response.body, $response.headers, $response.statusCode

The cookbook doesn't cover cursor APIs, but in our experience they (?cursor=abc) fit the first pattern with Value set to the cursor from the previous response, for example {{ $response.body.next_cursor }}. Check the API's docs for how the last page is signalled, and test with a small page size first so you can see the stop condition work.

Rate limits and retries

A 429 from the API means you are sending too fast. Two settings fix most cases (common issues):

• Batching: Items per Batch 1, Batch Interval 1000 sends one request per second.
• Retry On Fail (node Settings tab): set Max Tries and Wait Between Tries (ms).

For flaky APIs, add an error output too: Settings > On Error > Continue (using error output) sends failed items to a second branch instead of stopping the run.

Common errors

Error	Usual cause	Fix
Bad request - please check your parameters	Invalid query value or array format	Check Array Format in Query Parameters
The resource you are requesting could not be found	Wrong URL or deprecated endpoint	Compare with the API docs
ECONNREFUSED	localhost inside Docker points at the n8n container	Use host.docker.internal or the Compose service name
JSON parameter need to be an valid JSON	Broken JSON or expression	Wrap the whole object in {{ }}
Forbidden - perhaps check your credentials	403, missing scope	Regenerate the key with the right scopes
415 Unsupported media type	n8n Binary File sent where multipart was expected	Use Form-Data with a parameter of type n8n Binary File

On n8n Cloud there is no localhost: expose a local service through a tunnel if you need to call it during development.

HTTP Request node or Code node?

The n8n Code node docs say the Code node can't make HTTP requests and point you to this node. In n8n 2.x the Code node sandbox also has no fetch. So the default answer is: request in the HTTP Request node, transform in the Code node.

That split also gets you credentials management, retries, batching and pagination for free. Code that only prepared a payload before calling an API becomes an expression or a small Code node in front of the HTTP Request node.

Moving HTTP calls from Pipedream

Pipedream workflows call APIs in two ways: the Send any HTTP Request action and axios inside code steps (Pipedream HTTP docs). The action maps cleanly to this node. The axios calls are buried in code, which is where migrations slow down.

Our converter turns HTTP actions into HTTP Request nodes and lists every code step that calls an API, so you know which steps to split. Check a single code step for free with the code step checker, or run your whole export through the free report. The report and 1 workflow are free. Full export: $3 per workflow, $19 minimum — coming soon. The full walkthrough is in migrate Pipedream to n8n.

Not affiliated with Pipedream or n8n.