Ky is a tiny and elegant HTTP client based on the Fetch API

Ky targets modern browsers, Node.js, Bun, and Deno.

It's just a tiny package with no dependencies.

Benefits over plain fetch

• Simpler API
• Method shortcuts (ky.post())
• Treats non-2xx status codes as errors (after redirects)
• Retries failed requests
• JSON option
• Timeout support
• URL prefix option
• Instances with custom defaults
• Hooks
• TypeScript niceties (e.g. .json() supports generics and defaults to unknown, not any)

Install

npm install ky

CDN

• jsdelivr
• unpkg
• esm.sh

Usage

import ky from 'ky';

const json = await ky.post('https://example.com', {json: {foo: true}}).json();

console.log(json);
//=> `{data: '🦄'}`

With plain fetch, it would be:

class HTTPError extends Error {}

const response = await fetch('https://example.com', {
	method: 'POST',
	body: JSON.stringify({foo: true}),
	headers: {
		'content-type': 'application/json'
	}
});

if (!response.ok) {
	throw new HTTPError(`Fetch error: ${response.statusText}`);
}

const json = await response.json();

console.log(json);
//=> `{data: '🦄'}`

If you are using Deno, import Ky from a URL. For example, using a CDN:

import ky from 'https://esm.sh/ky';

API

ky(input, options?)

The input and options are the same as fetch, with additional options available (see below).

Returns a Response object with Body methods added for convenience. So you can, for example, call ky.get(input).json() directly without having to await the Response first. When called like that, an appropriate Accept header will be set depending on the body method used. Unlike the Body methods of window.Fetch; these will throw an HTTPError if the response status is not in the range of 200...299. Also, .json() will return an empty string if body is empty or the response status is 204 instead of throwing a parse error due to an empty body.

import ky from 'ky';

const user = await ky('/api/user').json();

console.log(user);

⌨️ TypeScript: Accepts an optional type parameter, which defaults to unknown, and is passed through to the return type of .json().

import ky from 'ky';

// user1 is unknown
const user1 = await ky('/api/users/1').json();
// user2 is a User
const user2 = await ky<User>('/api/users/2').json();
// user3 is a User
const user3 = await ky('/api/users/3').json<User>();

console.log([user1, user2, user3]);

ky.get(input, options?)

ky.post(input, options?)

ky.put(input, options?)

ky.patch(input, options?)

ky.head(input, options?)

ky.delete(input, options?)

Sets options.method to the method name and makes a request.

⌨️ TypeScript: Accepts an optional type parameter for use with JSON responses (see ky()).

input

Type: string | URL | Request

Same as fetch input.

When using a Request instance as input, any URL altering options (such as prefixUrl) will be ignored.

options

Type: object

Same as fetch options, plus the following additional options:

method

Type: string
Default: 'get'

HTTP method used to make the request.

Internally, the standard methods (GET, POST, PUT, PATCH, HEAD and DELETE) are uppercased in order to avoid server errors due to case sensitivity.

json

Type: object and any other value accepted by JSON.stringify()

Shortcut for sending JSON. Use this instead of the body option. Accepts any plain object or value, which will be JSON.stringify()'d and sent in the body with the correct header set.

searchParams

Type: string | object<string, string | number | boolean> | Array<Array<string | number | boolean>> | URLSearchParams
Default: ''

Search parameters to include in the request URL. Setting this will override all existing search parameters in the input URL.

Accepts any value supported by URLSearchParams().

prefixUrl

Type: string | URL

A prefix to prepend to the input URL when making the request. It can be any valid URL, either relative or absolute. A trailing slash / is optional and will be added automatically, if needed, when it is joined with input. Only takes effect when input is a string. The input argument cannot start with a slash / when using this option.

Useful when used with ky.extend() to create niche-specific Ky-instances.

import ky from 'ky';

// On https://example.com

const response = await ky('unicorn', {prefixUrl: '/api'});
//=> 'https://example.com/api/unicorn'

const response2 = await ky('unicorn', {prefixUrl: 'https://cats.com'});
//=> 'https://cats.com/unicorn'

Notes:

• After prefixUrl and input are joined, the result is resolved against the base URL of the page (if any).
• Leading slashes in input are disallowed when using this option to enforce consistency and avoid confusion about how the input URL is handled, given that input will not follow the normal URL resolution rules when prefixUrl is being used, which changes the meaning of a leading slash.

retry

Type: object | number
Default:

• limit: 2
• methods: get put head delete options trace
• statusCodes: 408 413 429 500 502 503 504
• afterStatusCodes: 413, 429, 503
• maxRetryAfter: undefined
• backoffLimit: undefined
• delay: attemptCount => 0.3 * (2 ** (attemptCount - 1)) * 1000

An object representing limit, methods, statusCodes, afterStatusCodes, and maxRetryAfter fields for maximum retry count, allowed methods, allowed status codes, status codes allowed to use the Retry-After time, and maximum Retry-After time.

If retry is a number, it will be used as limit and other defaults will remain in place.

If the response provides an HTTP status contained in afterStatusCodes, Ky will wait until the date, timeout, or timestamp given in the Retry-After header has passed to retry the request. If Retry-After is missing, the non-standard RateLimit-Reset header is used in its place as a fallback. If the provided status code is not in the list, the Retry-After header will be ignored.

If maxRetryAfter is set to undefined, it will use options.timeout. If Retry-After header is greater than maxRetryAfter, it will use maxRetryAfter.

The backoffLimit option is the upper limit of the delay per retry in milliseconds. To clamp the delay, set backoffLimit to 1000, for example. By default, the delay is calculated with 0.3 * (2 ** (attemptCount - 1)) * 1000. The delay increases exponentially.

The delay option can be used to change how the delay between retries is calculated. The function receives one parameter, the attempt count, starting at 1.

Retries are not triggered following a timeout.

import ky from 'ky';

const json = await ky('https://example.com', {
	retry: {
		limit: 10,
		methods: ['get'],
		statusCodes: [413],
		backoffLimit: 3000
	}
}).json();

timeout

Type: number | false
Default: 10000

Timeout in milliseconds for getting a response, including any retries. Can not be greater than 2147483647. If set to false, there will be no timeout.

hooks

Type: object<string, Function[]>
Default: {beforeRequest: [], beforeRetry: [], afterResponse: []}

Hooks allow modifications during the request lifecycle. Hook functions may be async and are run serially.

hooks.beforeRequest

Type: Function[]
Default: []

This hook enables you to modify the request right before it is sent. Ky will make no further changes to the request after this. The hook function receives request and options as arguments. You could, for example, modify the request.headers here.

The hook can return a Request to replace the outgoing request, or return a Response to completely avoid making an HTTP request. This can be used to mock a request, check an internal cache, etc. An important consideration when returning a request or response from this hook is that any remaining beforeRequest hooks will be skipped, so you may want to only return them from the last hook.

import ky from 'ky';

const api = ky.extend({
	hooks: {
		beforeRequest: [
			request => {
				request.headers.set('X-Requested-With', 'ky');
			}
		]
	}
});

const response = await api.get('https://example.com/api/users');

hooks.beforeRetry

Type: Function[]
Default: []

This hook enables you to modify the request right before retry. Ky will make no further changes to the request after this. The hook function receives an object with the normalized request and options, an error instance, and the retry count. You could, for example, modify request.headers here.

If the request received a response, the error will be of type HTTPError and the Response object will be available at error.response. Be aware that some types of errors, such as network errors, inherently mean that a response was not received. In that case, the error will not be an instance of HTTPError.

You can prevent Ky from retrying the request by throwing an error. Ky will not handle it in any way and the error will be propagated to the request initiator. The rest of the beforeRetry hooks will not be called in this case. Alternatively, you can return the ky.stop symbol to do the same thing but without propagating an error (this has some limitations, see ky.stop docs for details).

import ky from 'ky';

const response = await ky('https://example.com', {
	hooks: {
		beforeRetry: [
			async ({request, options, error, retryCount}) => {
				const token = await ky('https://example.com/refresh-token');
				request.headers.set('Authorization', `token ${token}`);
			}
		]
	}
});

hooks.beforeError

Type: Function[]
Default: []

This hook enables you to modify the HTTPError right before it is thrown. The hook function receives a HTTPError as an argument and should return an instance of HTTPError.

import ky from 'ky';

await ky('https://example.com', {
	hooks: {
		beforeError: [
			error => {
				const {response} = error;
				if (response && response.body) {
					error.name = 'GitHubError';
					error.message = `${response.body.message} (${response.status})`;
				}

				return error;
			}
		]
	}
});

hooks.afterResponse

Type: Function[]
Default: []

This hook enables you to read and optionally modify the response. The hook function receives normalized request, options, and a clone of the response as arguments. The return value of the hook function will be used by Ky as the response object if it's an instance of Response.

import ky from 'ky';

const response = await ky('https://example.com', {
	hooks: {
		afterResponse: [
			(_request, _options, response) => {
				// You could do something with the response, for example, logging.
				log(response);

				// Or return a `Response` instance to overwrite the response.
				return new Response('A different response', {status: 200});
			},

			// Or retry with a fresh token on a 403 error
			async (request, options, response) => {
				if (response.status === 403) {
					// Get a fresh token
					const token = await ky('https://example.com/token').text();

					// Retry with the token
					request.headers.set('Authorization', `token ${token}`);

					return ky(request);
				}
			}
		]
	}
});

throwHttpErrors

Type: boolean
Default: true

Throw an HTTPError when, after following redirects, the response has a non-2xx status code. To also throw for redirects instead of following them, set the redirect option to 'manual'.

Setting this to false may be useful if you are checking for resource availability and are expecting error responses.

Note: If false, error responses are considered successful and the request will not be retried.

onDownloadProgress

Type: Function

Download progress event handler.

The function receives a progress and chunk argument:

• The progress object contains the following elements: percent, transferredBytes and totalBytes. If it's not possible to retrieve the body size, totalBytes will be 0.
• The chunk argument is an instance of Uint8Array. It's empty for the first call.

import ky from 'ky';

const response = await ky('https://example.com', {
	onDownloadProgress: (progress, chunk) => {
		// Example output:
		// `0% - 0 of 1271 bytes`
		// `100% - 1271 of 1271 bytes`
		console.log(`${progress.percent * 100}% - ${progress.transferredBytes} of ${progress.totalBytes} bytes`);
	}
});

parseJson

Type: Function
Default: JSON.parse()

User-defined JSON-parsing function.

Use-cases:

1. Parse JSON via the bourne package to protect from prototype pollution.
2. Parse JSON with reviver option of JSON.parse().

import ky from 'ky';
import bourne from '@hapijs/bourne';

const json = await ky('https://example.com', {
	parseJson: text => bourne(text)
}).json();

stringifyJson

Type: Function
Default: JSON.stringify()

User-defined JSON-stringifying function.

Use-cases:

1. Stringify JSON with a custom replacer function.

import ky from 'ky';
import {DateTime} from 'luxon';

const json = await ky('https://example.com', {
	stringifyJson: data => JSON.stringify(data, (key, value) => {
		if (key.endsWith('_at')) {
			return DateTime.fromISO(value).toSeconds();
		}

		return value;
	})
}).json();

fetch

Type: Function
Default: fetch

User-defined fetch function. Has to be fully compatible with the Fetch API standard.

Use-cases:

1. Use custom fetch implementations like isomorphic-unfetch.
2. Use the fetch wrapper function provided by some frameworks that use server-side rendering (SSR).

import ky from 'ky';
import fetch from 'isomorphic-unfetch';

const json = await ky('https://example.com', {fetch}).json();

ky.extend(defaultOptions)

Create a new ky instance with some defaults overridden with your own.

In contrast to ky.create(), ky.extend() inherits defaults from its parent.

You can pass headers as a Headers instance or a plain object.

You can remove a header with .extend() by passing the header with an undefined value. Passing undefined as a string removes the header only if it comes from a Headers instance.

Similarly, you can remove existing hooks entries by extending the hook with an explicit undefined.

import ky from 'ky';

const url = 'https://sindresorhus.com';

const original = ky.create({
	headers: {
		rainbow: 'rainbow',
		unicorn: 'unicorn'
	},
	hooks: {
		beforeRequest: [ () => console.log('before 1') ],
		afterResponse: [ () => console.log('after 1') ],
	},
});

const extended = original.extend({
	headers: {
		rainbow: undefined
	},
	hooks: {
		beforeRequest: undefined,
		afterResponse: [ () => console.log('after 2') ],
	}
});

const response = await extended(url).json();
//=> after 1
//=> after 2

console.log('rainbow' in response);
//=> false

console.log('unicorn' in response);
//=> true

You can also refer to parent defaults by providing a function to .extend().

import ky from 'ky';

const api = ky.create({prefixUrl: 'https://example.com/api'});

const usersApi = api.extend((options) => ({prefixUrl: `${options.prefixUrl}/users`}));

const response = await usersApi.get('123');
//=> 'https://example.com/api/users/123'

const response = await api.get('version');
//=> 'https://example.com/api/version'

ky.create(defaultOptions)

Create a new Ky instance with complete new defaults.

import ky from 'ky';

// On https://my-site.com

const api = ky.create({prefixUrl: 'https://example.com/api'});

const response = await api.get('users/123');
//=> 'https://example.com/api/users/123'

const response = await api.get('/status', {prefixUrl: ''});
//=> 'https://my-site.com/status'

defaultOptions

Type: object

ky.stop

A Symbol that can be returned by a beforeRetry hook to stop the retry. This will also short circuit the remaining beforeRetry hooks.

Note: Returning this symbol makes Ky abort and return with an undefined response. Be sure to check for a response before accessing any properties on it or use optional chaining. It is also incompatible with body methods, such as .json() or .text(), because there is no response to parse. In general, we recommend throwing an error instead of returning this symbol, as that will cause Ky to abort and then throw, which avoids these limitations.

A valid use-case for ky.stop is to prevent retries when making requests for side effects, where the returned data is not important. For example, logging client activity to the server.

import ky from 'ky';

const options = {
	hooks: {
		beforeRetry: [
			async ({request, options, error, retryCount}) => {
				const shouldStopRetry = await ky('https://example.com/api');
				if (shouldStopRetry) {
					return ky.stop;
				}
			}
		]
	}
};

// Note that response will be `undefined` in case `ky.stop` is returned.
const response = await ky.post('https://example.com', options);

// Using `.text()` or other body methods is not supported.
const text = await ky('https://example.com', options).text();

HTTPError

Exposed for instanceof checks. The error has a response property with the Response object, request property with the Request object, and options property with normalized options (either passed to ky when creating an instance with ky.create() or directly when performing the request).

If you need to read the actual response when an HTTPError has occurred, call the respective parser method on the response object. For example:

try {
	await ky('https://example.com').json();
} catch (error) {
	if (error.name === 'HTTPError') {
		const errorJson = await error.response.json();
	}
}

⌨️ TypeScript: Accepts an optional type parameter, which defaults to unknown, and is passed through to the return type of error.response.json().

TimeoutError

The error thrown when the request times out. It has a request property with the Request object.

Tips

Sending form data

Sending form data in Ky is identical to fetch. Just pass a FormData instance to the body option. The Content-Type header will be automatically set to multipart/form-data.

import ky from 'ky';

// `multipart/form-data`
const formData = new FormData();
formData.append('food', 'fries');
formData.append('drink', 'icetea');

const response = await ky.post(url, {body: formData});

If you want to send the data in application/x-www-form-urlencoded format, you will need to encode the data with URLSearchParams.

import ky from 'ky';

// `application/x-www-form-urlencoded`
const searchParams = new URLSearchParams();
searchParams.set('food', 'fries');
searchParams.set('drink', 'icetea');

const response = await ky.post(url, {body: searchParams});

Setting a custom Content-Type

Ky automatically sets an appropriate Content-Type header for each request based on the data in the request body. However, some APIs require custom, non-standard content types, such as application/x-amz-json-1.1. Using the headers option, you can manually override the content type.

import ky from 'ky';

const json = await ky.post('https://example.com', {
	headers: {
		'content-type': 'application/json'
	},
	json: {
		foo: true
	},
}).json();

console.log(json);
//=> `{data: '🦄'}`

Cancellation

Fetch (and hence Ky) has built-in support for request cancellation through the AbortController API. Read more.

Example:

import ky from 'ky';

const controller = new AbortController();
const {signal} = controller;

setTimeout(() => {
	controller.abort();
}, 5000);

try {
	console.log(await ky(url, {signal}).text());
} catch (error) {
	if (error.name === 'AbortError') {
		console.log('Fetch aborted');
	} else {
		console.error('Fetch error:', error);
	}
}

FAQ

How do I use this in Node.js?

Node.js 18 and later supports fetch natively, so you can just use this package directly.

How do I use this with a web app (React, Vue.js, etc.) that uses server-side rendering (SSR)?

Same as above.

How do I test a browser library that uses this?

Either use a test runner that can run in the browser, like Mocha, or use AVA with ky-universal. Read more.

How do I use this without a bundler like Webpack?

Make sure your code is running as a JavaScript module (ESM), for example by using a <script type="module"> tag in your HTML document. Then Ky can be imported directly by that module without a bundler or other tools.

<script type="module">
import ky from 'https://unpkg.com/ky/distribution/index.js';

const json = await ky('https://jsonplaceholder.typicode.com/todos/1').json();

console.log(json.title);
//=> 'delectus aut autem'
</script>

How is it different from got

See my answer here. Got is maintained by the same people as Ky.

How is it different from axios?

See my answer here.

How is it different from r2?

See my answer in #10.

What does ky mean?

It's just a random short npm package name I managed to get. It does, however, have a meaning in Japanese:

A form of text-able slang, KY is an abbreviation for 空気読めない (kuuki yomenai), which literally translates into “cannot read the air.” It's a phrase applied to someone who misses the implied meaning.

Browser support

The latest version of Chrome, Firefox, and Safari.

Node.js support

Node.js 18 and later.

Related

• fetch-extras - Useful utilities for working with Fetch
• got - Simplified HTTP requests for Node.js
• ky-hooks-change-case - Ky hooks to modify cases on requests and responses of objects

Maintainers

• Sindre Sorhus
• Szymon Marczak
• Seth Holladay