DefinitelyTyped icon, indicating that this package has TypeScript declarations provided by the separate @types/tough-cookie package

    4.0.0 • Public • Published

    RFC6265 Cookies and CookieJar for Node.js

    npm package

    Build Status


    var tough = require('tough-cookie');
    var Cookie = tough.Cookie;
    var cookie = Cookie.parse(header);
    cookie.value = 'somethingdifferent';
    header = cookie.toString();
    var cookiejar = new tough.CookieJar();
    cookiejar.setCookie(cookie, '', cb);
    // ...
    cookiejar.getCookies('',function(err,cookies) {
      res.headers['cookie'] = cookies.join('');


    It's so easy!

    npm install tough-cookie

    Why the name? NPM modules cookie, cookies and cookiejar were already taken.

    Version Support

    Support for versions of node.js will follow that of the request module.



    Functions on the module you get from require('tough-cookie'). All can be used as pure functions and don't need to be "bound".

    Note: prior to 1.0.x, several of these functions took a strict parameter. This has since been removed from the API as it was no longer necessary.


    Parse a cookie date string into a Date. Parses according to RFC6265 Section 5.1.1, not Date.parse().


    Format a Date into a RFC1123 string (the RFC6265-recommended format).


    Transforms a domain-name into a canonical domain-name. The canonical domain-name is a trimmed, lowercased, stripped-of-leading-dot and optionally punycode-encoded domain-name (Section 5.1.2 of RFC6265). For the most part, this function is idempotent (can be run again on its output without ill effects).


    Answers "does this real domain match the domain in a cookie?". The str is the "current" domain-name and the domStr is the "cookie" domain-name. Matches according to RFC6265 Section 5.1.3, but it helps to think of it as a "suffix match".

    The canonicalize parameter will run the other two parameters through canonicalDomain or not.


    Given a current request/response path, gives the Path apropriate for storing in a cookie. This is basically the "directory" of a "file" in the path, but is specified by Section 5.1.4 of the RFC.

    The path parameter MUST be only the pathname part of a URI (i.e. excludes the hostname, query, fragment, etc.). This is the .pathname property of node's uri.parse() output.


    Answers "does the request-path path-match a given cookie-path?" as per RFC6265 Section 5.1.4. Returns a boolean.

    This is essentially a prefix-match where cookiePath is a prefix of reqPath.

    parse(cookieString[, options])

    alias for Cookie.parse(cookieString[, options])


    alias for Cookie.fromJSON(string)


    Returns the public suffix of this hostname. The public suffix is the shortest domain-name upon which a cookie can be set. Returns null if the hostname cannot have cookies set for it.

    For example: and both have public suffix

    For further information, see This module derives its list from that site. This call is currently a wrapper around psl's get() method.


    For use with .sort(), sorts a list of cookies into the recommended order given in the RFC (Section 5.4 step 2). The sort algorithm is, in order of precedence:

    • Longest .path
    • oldest .creation (which has a 1ms precision, same as Date)
    • lowest .creationIndex (to get beyond the 1ms precision)
    var cookies = [ /* unsorted array of Cookie objects */ ];
    cookies = cookies.sort(cookieCompare);

    Note: Since JavaScript's Date is limited to a 1ms precision, cookies within the same milisecond are entirely possible. This is especially true when using the now option to .setCookie(). The .creationIndex property is a per-process global counter, assigned during construction with new Cookie(). This preserves the spirit of the RFC sorting: older cookies go first. This works great for MemoryCookieStore, since Set-Cookie headers are parsed in order, but may not be so great for distributed systems. Sophisticated Stores may wish to set this to some other logical clock such that if cookies A and B are created in the same millisecond, but cookie A is created before cookie B, then A.creationIndex < B.creationIndex. If you want to alter the global counter, which you probably shouldn't do, it's stored in Cookie.cookiesCreated.


    Generates a list of all possible domains that domainMatch() the parameter. May be handy for implementing cookie stores.


    Generates a list of all possible paths that pathMatch() the parameter. May be handy for implementing cookie stores.


    Exported via tough.Cookie.

    Cookie.parse(cookieString[, options])

    Parses a single Cookie or Set-Cookie HTTP header into a Cookie object. Returns undefined if the string can't be parsed.

    The options parameter is not required and currently has only one property:

    • loose - boolean - if true enable parsing of key-less cookies like =abc and =, which are not RFC-compliant.

    If options is not an object, it is ignored, which means you can use Array#map with it.

    Here's how to process the Set-Cookie header(s) on a node HTTP/HTTPS response:

    if (res.headers['set-cookie'] instanceof Array)
      cookies = res.headers['set-cookie'].map(Cookie.parse);
      cookies = [Cookie.parse(res.headers['set-cookie'])];

    Note: in version 2.3.3, tough-cookie limited the number of spaces before the = to 256 characters. This limitation has since been removed. See Issue 92


    Cookie object properties:

    • key - string - the name or key of the cookie (default "")
    • value - string - the value of the cookie (default "")
    • expires - Date - if set, the Expires= attribute of the cookie (defaults to the string "Infinity"). See setExpires()
    • maxAge - seconds - if set, the Max-Age= attribute in seconds of the cookie. May also be set to strings "Infinity" and "-Infinity" for non-expiry and immediate-expiry, respectively. See setMaxAge()
    • domain - string - the Domain= attribute of the cookie
    • path - string - the Path= of the cookie
    • secure - boolean - the Secure cookie flag
    • httpOnly - boolean - the HttpOnly cookie flag
    • sameSite - string - the SameSite cookie attribute (from [RFC6265bis]); must be one of none, lax, or strict
    • extensions - Array - any unrecognized cookie attributes as strings (even if equal-signs inside)
    • creation - Date - when this cookie was constructed
    • creationIndex - number - set at construction, used to provide greater sort precision (please see cookieCompare(a,b) for a full explanation)

    After a cookie has been passed through CookieJar.setCookie() it will have the following additional attributes:

    • hostOnly - boolean - is this a host-only cookie (i.e. no Domain field was set, but was instead implied)
    • pathIsDefault - boolean - if true, there was no Path field on the cookie and defaultPath() was used to derive one.
    • creation - Date - modified from construction to when the cookie was added to the jar
    • lastAccessed - Date - last time the cookie got accessed. Will affect cookie cleaning once implemented. Using cookiejar.getCookies(...) will update this attribute.


    Receives an options object that can contain any of the above Cookie properties, uses the default for unspecified properties.


    encode to a Set-Cookie header value. The Expires cookie field is set using formatDate(), but is omitted entirely if .expires is Infinity.


    encode to a Cookie header value (i.e. the .key and .value properties joined with '=').


    sets the expiry based on a date-string passed through parseDate(). If parseDate returns null (i.e. can't parse this date string), .expires is set to "Infinity" (a string) is set.


    sets the maxAge in seconds. Coerces -Infinity to "-Infinity" and Infinity to "Infinity" so it JSON serializes correctly.



    expiryTime() Computes the absolute unix-epoch milliseconds that this cookie expires. expiryDate() works similarly, except it returns a Date object. Note that in both cases the now parameter should be milliseconds.

    Max-Age takes precedence over Expires (as per the RFC). The .creation attribute -- or, by default, the now parameter -- is used to offset the .maxAge attribute.

    If Expires (.expires) is set, that's returned.

    Otherwise, expiryTime() returns Infinity and expiryDate() returns a Date object for "Tue, 19 Jan 2038 03:14:07 GMT" (latest date that can be expressed by a 32-bit time_t; the common limit for most user-agents).


    compute the TTL relative to now (milliseconds). The same precedence rules as for expiryTime/expiryDate apply.

    The "number" Infinity is returned for cookies without an explicit expiry and 0 is returned if the cookie is expired. Otherwise a time-to-live in milliseconds is returned.



    return the canonicalized .domain field. This is lower-cased and punycode (RFC3490) encoded if the domain has any non-ASCII characters.


    For convenience in using JSON.serialize(cookie). Returns a plain-old Object that can be JSON-serialized.

    Any Date properties (i.e., .expires, .creation, and .lastAccessed) are exported in ISO format (.toISOString()).

    NOTE: Custom Cookie properties will be discarded. In tough-cookie 1.x, since there was no .toJSON method explicitly defined, all enumerable properties were captured. If you want a property to be serialized, add the property name to the Cookie.serializableProperties Array.


    Does the reverse of cookie.toJSON(). If passed a string, will JSON.parse() that first.

    Any Date properties (i.e., .expires, .creation, and .lastAccessed) are parsed via Date.parse(), not the tough-cookie parseDate, since it's JavaScript/JSON-y timestamps being handled at this layer.

    Returns null upon JSON parsing error.


    Does a deep clone of this cookie, exactly implemented as Cookie.fromJSON(cookie.toJSON()).


    Status: IN PROGRESS. Works for a few things, but is by no means comprehensive.

    validates cookie attributes for semantic correctness. Useful for "lint" checking any Set-Cookie headers you generate. For now, it returns a boolean, but eventually could return a reason string -- you can future-proof with this construct:

    if (cookie.validate() === true) {
      // it's tasty
    } else {
      // yuck!


    Exported via tough.CookieJar.


    Simply use new CookieJar(). If you'd like to use a custom store, pass that to the constructor otherwise a MemoryCookieStore will be created and used.

    The options object can be omitted and can have the following properties:

    • rejectPublicSuffixes - boolean - default true - reject cookies with domains like "com" and ""
    • looseMode - boolean - default false - accept malformed cookies like bar and =bar, which have an implied empty name.
    • prefixSecurity - string - default silent - set to 'unsafe-disabled', 'silent', or 'strict'. See [Cookie Prefixes] below.
    • allowSpecialUseDomain - boolean - default false - accepts special-use domain suffixes, such as local. Useful for testing purposes. This is not in the standard, but is used sometimes on the web and is accepted by (most) browsers.

    Since eventually this module would like to support database/remote/etc. CookieJars, continuation passing style is used for CookieJar methods.

    .setCookie(cookieOrString, currentUrl, [{options},] cb(err,cookie))

    Attempt to set the cookie in the cookie jar. If the operation fails, an error will be given to the callback cb, otherwise the cookie is passed through. The cookie will have updated .creation, .lastAccessed and .hostOnly properties.

    The options object can be omitted and can have the following properties:

    • http - boolean - default true - indicates if this is an HTTP or non-HTTP API. Affects HttpOnly cookies.
    • secure - boolean - autodetect from url - indicates if this is a "Secure" API. If the currentUrl starts with https: or wss: then this is defaulted to true, otherwise false.
    • now - Date - default new Date() - what to use for the creation/access time of cookies
    • ignoreError - boolean - default false - silently ignore things like parse errors and invalid domains. Store errors aren't ignored by this option.
    • sameSiteContext - string - default unset - set to 'none', 'lax', or 'strict' See [SameSite Cookies] below.

    As per the RFC, the .hostOnly property is set if there was no "Domain=" parameter in the cookie string (or .domain was null on the Cookie object). The .domain property is set to the fully-qualified hostname of currentUrl in this case. Matching this cookie requires an exact hostname match (not a domainMatch as per usual).

    .setCookieSync(cookieOrString, currentUrl, [{options}])

    Synchronous version of setCookie; only works with synchronous stores (e.g. the default MemoryCookieStore).

    .getCookies(currentUrl, [{options},] cb(err,cookies))

    Retrieve the list of cookies that can be sent in a Cookie header for the current url.

    If an error is encountered, that's passed as err to the callback, otherwise an Array of Cookie objects is passed. The array is sorted with cookieCompare() unless the {sort:false} option is given.

    The options object can be omitted and can have the following properties:

    • http - boolean - default true - indicates if this is an HTTP or non-HTTP API. Affects HttpOnly cookies.
    • secure - boolean - autodetect from url - indicates if this is a "Secure" API. If the currentUrl starts with https: or wss: then this is defaulted to true, otherwise false.
    • now - Date - default new Date() - what to use for the creation/access time of cookies
    • expire - boolean - default true - perform expiry-time checking of cookies and asynchronously remove expired cookies from the store. Using false will return expired cookies and not remove them from the store (which is useful for replaying Set-Cookie headers, potentially).
    • allPaths - boolean - default false - if true, do not scope cookies by path. The default uses RFC-compliant path scoping. Note: may not be supported by the underlying store (the default MemoryCookieStore supports it).
    • sameSiteContext - string - default unset - Set this to 'none', 'lax' or 'strict' to enforce SameSite cookies upon retrival. See [SameSite Cookies] below.

    The .lastAccessed property of the returned cookies will have been updated.

    .getCookiesSync(currentUrl, [{options}])

    Synchronous version of getCookies; only works with synchronous stores (e.g. the default MemoryCookieStore).


    Accepts the same options as .getCookies() but passes a string suitable for a Cookie header rather than an array to the callback. Simply maps the Cookie array via .cookieString().


    Synchronous version of getCookieString; only works with synchronous stores (e.g. the default MemoryCookieStore).


    Returns an array of strings suitable for Set-Cookie headers. Accepts the same options as .getCookies(). Simply maps the cookie array via .toString().


    Synchronous version of getSetCookieStrings; only works with synchronous stores (e.g. the default MemoryCookieStore).


    Serialize the Jar if the underlying store supports .getAllCookies.

    NOTE: Custom Cookie properties will be discarded. If you want a property to be serialized, add the property name to the Cookie.serializableProperties Array.

    See [Serialization Format].


    Sync version of .serialize


    Alias of .serializeSync() for the convenience of JSON.stringify(cookiejar).

    CookieJar.deserialize(serialized, [store], cb(err,object))

    A new Jar is created and the serialized Cookies are added to the underlying store. Each Cookie is added via store.putCookie in the order in which they appear in the serialization.

    The store argument is optional, but should be an instance of Store. By default, a new instance of MemoryCookieStore is created.

    As a convenience, if serialized is a string, it is passed through JSON.parse first. If that throws an error, this is passed to the callback.

    CookieJar.deserializeSync(serialized, [store])

    Sync version of .deserialize. Note that the store must be synchronous for this to work.


    Alias of .deserializeSync to provide consistency with Cookie.fromJSON().


    Produces a deep clone of this jar. Modifications to the original won't affect the clone, and vice versa.

    The store argument is optional, but should be an instance of Store. By default, a new instance of MemoryCookieStore is created. Transferring between store types is supported so long as the source implements .getAllCookies() and the destination implements .putCookie().


    Synchronous version of .clone, returning a new CookieJar instance.

    The store argument is optional, but must be a synchronous Store instance if specified. If not passed, a new instance of MemoryCookieStore is used.

    The source and destination must both be synchronous Stores. If one or both stores are asynchronous, use .clone instead. Recall that MemoryCookieStore supports both synchronous and asynchronous API calls.


    Removes all cookies from the jar.

    This is a new backwards-compatible feature of tough-cookie version 2.5, so not all Stores will implement it efficiently. For Stores that do not implement removeAllCookies, the fallback is to call removeCookie after getAllCookies. If getAllCookies fails or isn't implemented in the Store, that error is returned. If one or more of the removeCookie calls fail, only the first error is returned.


    Sync version of .removeAllCookies()


    Base class for CookieJar stores. Available as tough.Store.

    Store API

    The storage model for each CookieJar instance can be replaced with a custom implementation. The default is MemoryCookieStore which can be found in the lib/memstore.js file. The API uses continuation-passing-style to allow for asynchronous stores.

    Stores should inherit from the base Store class, which is available as require('tough-cookie').Store.

    Stores are asynchronous by default, but if store.synchronous is set to true, then the *Sync methods on the of the containing CookieJar can be used (however, the continuation-passing style

    All domain parameters will have been normalized before calling.

    The Cookie store must have all of the following methods.

    store.findCookie(domain, path, key, cb(err,cookie))

    Retrieve a cookie with the given domain, path and key (a.k.a. name). The RFC maintains that exactly one of these cookies should exist in a store. If the store is using versioning, this means that the latest/newest such cookie should be returned.

    Callback takes an error and the resulting Cookie object. If no cookie is found then null MUST be passed instead (i.e. not an error).

    store.findCookies(domain, path, cb(err,cookies))

    Locates cookies matching the given domain and path. This is most often called in the context of cookiejar.getCookies() above.

    If no cookies are found, the callback MUST be passed an empty array.

    The resulting list will be checked for applicability to the current request according to the RFC (domain-match, path-match, http-only-flag, secure-flag, expiry, etc.), so it's OK to use an optimistic search algorithm when implementing this method. However, the search algorithm used SHOULD try to find cookies that domainMatch() the domain and pathMatch() the path in order to limit the amount of checking that needs to be done.

    As of version 0.9.12, the allPaths option to cookiejar.getCookies() above will cause the path here to be null. If the path is null, path-matching MUST NOT be performed (i.e. domain-matching only).

    store.putCookie(cookie, cb(err))

    Adds a new cookie to the store. The implementation SHOULD replace any existing cookie with the same .domain, .path, and .key properties -- depending on the nature of the implementation, it's possible that between the call to fetchCookie and putCookie that a duplicate putCookie can occur.

    The cookie object MUST NOT be modified; the caller will have already updated the .creation and .lastAccessed properties.

    Pass an error if the cookie cannot be stored.

    store.updateCookie(oldCookie, newCookie, cb(err))

    Update an existing cookie. The implementation MUST update the .value for a cookie with the same domain, .path and .key. The implementation SHOULD check that the old value in the store is equivalent to oldCookie - how the conflict is resolved is up to the store.

    The .lastAccessed property will always be different between the two objects (to the precision possible via JavaScript's clock). Both .creation and .creationIndex are guaranteed to be the same. Stores MAY ignore or defer the .lastAccessed change at the cost of affecting how cookies are selected for automatic deletion (e.g., least-recently-used, which is up to the store to implement).

    Stores may wish to optimize changing the .value of the cookie in the store versus storing a new cookie. If the implementation doesn't define this method a stub that calls putCookie(newCookie,cb) will be added to the store object.

    The newCookie and oldCookie objects MUST NOT be modified.

    Pass an error if the newCookie cannot be stored.

    store.removeCookie(domain, path, key, cb(err))

    Remove a cookie from the store (see notes on findCookie about the uniqueness constraint).

    The implementation MUST NOT pass an error if the cookie doesn't exist; only pass an error due to the failure to remove an existing cookie.

    store.removeCookies(domain, path, cb(err))

    Removes matching cookies from the store. The path parameter is optional, and if missing means all paths in a domain should be removed.

    Pass an error ONLY if removing any existing cookies failed.


    Optional. Removes all cookies from the store.

    Pass an error if one or more cookies can't be removed.

    Note: New method as of tough-cookie version 2.5, so not all Stores will implement this, plus some stores may choose not to implement this.

    store.getAllCookies(cb(err, cookies))

    Optional. Produces an Array of all cookies during jar.serialize(). The items in the array can be true Cookie objects or generic Objects with the [Serialization Format] data structure.

    Cookies SHOULD be returned in creation order to preserve sorting via compareCookies(). For reference, MemoryCookieStore will sort by .creationIndex since it uses true Cookie objects internally. If you don't return the cookies in creation order, they'll still be sorted by creation time, but this only has a precision of 1ms. See compareCookies for more detail.

    Pass an error if retrieval fails.

    Note: not all Stores can implement this due to technical limitations, so it is optional.


    Inherits from Store.

    A just-in-memory CookieJar synchronous store implementation, used by default. Despite being a synchronous implementation, it's usable with both the synchronous and asynchronous forms of the CookieJar API. Supports serialization, getAllCookies, and removeAllCookies.

    Community Cookie Stores

    These are some Store implementations authored and maintained by the community. They aren't official and we don't vouch for them but you may be interested to have a look:

    Serialization Format

    NOTE: if you want to have custom Cookie properties serialized, add the property name to Cookie.serializableProperties.

        // The version of tough-cookie that serialized this jar.
        version: 'tough-cookie@1.x.y',
        // add the store type, to make humans happy:
        storeType: 'MemoryCookieStore',
        // CookieJar configuration:
        rejectPublicSuffixes: true,
        // ... future items go here
        // Gets filled from
        cookies: [
            key: 'string',
            value: 'string',
            // ...
            /* other Cookie.serializableProperties go here */


    Support for RFC6265bis revision 02 is being developed. Since this is a bit of an omnibus revision to the RFC6252, support is broken up into the functional areas.

    Leave Secure Cookies Alone

    Not yet supported.

    This change makes it so that if a cookie is sent from the server to the client with a Secure attribute, the channel must also be secure or the cookie is ignored.

    SameSite Cookies


    This change makes it possible for servers, and supporting clients, to mitigate certain types of CSRF attacks by disallowing SameSite cookies from being sent cross-origin.

    On the Cookie object itself, you can get/set the .sameSite attribute, which will be serialized into the SameSite= cookie attribute. When unset or undefined, no SameSite= attribute will be serialized. The valid values of this attribute are 'none', 'lax', or 'strict'. Other values will be serialized as-is.

    When parsing cookies with a SameSite cookie attribute, values other than 'lax' or 'strict' are parsed as 'none'. For example, SomeCookie=SomeValue; SameSite=garbage will parse so that cookie.sameSite === 'none'.

    In order to support SameSite cookies, you must provide a sameSiteContext option to both setCookie and getCookies. Valid values for this option are just like for the Cookie object, but have particular meanings:

    1. 'strict' mode - If the request is on the same "site for cookies" (see the RFC draft for what this means), pass this option to add a layer of defense against CSRF.
    2. 'lax' mode - If the request is from another site, but is directly because of navigation by the user, e.g., <link type=prefetch> or <a href="...">, pass sameSiteContext: 'lax'.
    3. 'none' - Otherwise, pass sameSiteContext: 'none' (this indicates a cross-origin request).
    4. unset/undefined - SameSite will not be enforced! This can be a valid use-case for when CSRF isn't in the threat model of the system being built.

    It is highly recommended that you read RFC 6265bis for fine details on SameSite cookies. In particular Section 8.8 discusses security considerations and defense in depth.

    Cookie Prefixes


    Cookie prefixes are a way to indicate that a given cookie was set with a set of attributes simply by inspecting the first few characters of the cookie's name.

    Cookie prefixes are defined in Section 4.1.3 of 6265bis. Two prefixes are defined:

    1. "__Secure-" Prefix: If a cookie's name begins with a case-sensitive match for the string "__Secure-", then the cookie will have been set with a "Secure" attribute.
    2. "__Host-" Prefix: If a cookie's name begins with a case-sensitive match for the string "__Host-", then the cookie will have been set with a "Secure" attribute, a "Path" attribute with a value of "/", and no "Domain" attribute.

    If prefixSecurity is enabled for CookieJar, then cookies that match the prefixes defined above but do not obey the attribute restrictions will not be added.

    You can define this functionality by passing in prefixSecurity option to CookieJar. It can be one of 3 values:

    1. silent: Enable cookie prefix checking but silently fail to add the cookie if conditions not met. Default.
    2. strict: Enable cookie prefix checking and error out if conditions not met.
    3. unsafe-disabled: Disable cookie prefix checking.

    Note that if ignoreError is passed in as true then the error will be silent regardless of prefixSecurity option (assuming it's enabled).

    Copyright and License


     Copyright (c) 2015,, Inc.
     All rights reserved.
     Redistribution and use in source and binary forms, with or without
     modification, are permitted provided that the following conditions are met:
     1. Redistributions of source code must retain the above copyright notice,
     this list of conditions and the following disclaimer.
     2. Redistributions in binary form must reproduce the above copyright notice,
     this list of conditions and the following disclaimer in the documentation
     and/or other materials provided with the distribution.
     3. Neither the name of nor the names of its contributors may
     be used to endorse or promote products derived from this software without
     specific prior written permission.


    npm i tough-cookie

    DownloadsWeekly Downloads






    Unpacked Size

    98.6 kB

    Total Files


    Last publish


    • awaterma
    • jstash
    • ruoho