openid-client is a server side OpenID Relying Party (RP, Client) implementation for Node.js, supports passport.
Table of Contents
- Implemented specs & features
- Certification
- Example
- Get started
- Usage
- Usage with passport
- Configuration
The following client/RP features from OpenID Connect/OAuth2.0 specifications are implemented by openid-client.
- OpenID Connect Core 1.0 incorporating errata set 1
- Authorization Callback
- Authorization Code Flow
- Implicit Flow
- Hybrid Flow
- UserInfo Request
- Fetching Distributed Claims
- Unpacking Aggregated Claims
- Offline Access / Refresh Token Grant
- Client Credentials Grant
- Password Grant
- Client Authentication
- client_secret_basic
- client_secret_post
- client_secret_jwt
- private_key_jwt
- Authorization Callback
- OpenID Connect Discovery 1.0 incorporating errata set 1
- Discovery of OpenID Provider (Issuer) Metadata
- Discovery of OpenID Provider (Issuer) Metadata via user provided inputs (see #WebFinger)
- OpenID Connect Dynamic Client Registration 1.0 incorporating errata set 1
- Dynamic Client Registration request
- Client initialization via registration client uri
- RFC7009 - OAuth 2.0 Token revocation
- Client Authenticated request to token revocation
- RFC7662 - OAuth 2.0 Token introspection
- Client Authenticated request to token introspection
Filip Skokan has certified that openid-client
conforms to the RP Basic, RP Implicit, RP Hybrid, RP Config and RP Dynamic profiles
of the OpenID Connect™ protocol.
Head over to the example folder to see the library in use. This example is deployed and configured to use an example OpenID Connect Provider here. The provider is using oidc-provider library.
On the off-chance you want to manage multiple clients for multiple issuers you need to first get an Issuer instance.
const Issuer = require('openid-client').Issuer;
Issuer.discover('https://accounts.google.com') // => Promise
.then(function (googleIssuer) {
console.log('Discovered issuer %s', googleIssuer);
});
const Issuer = require('openid-client').Issuer;
const googleIssuer = new Issuer({
issuer: 'https://accounts.google.com',
authorization_endpoint: 'https://accounts.google.com/o/oauth2/v2/auth',
token_endpoint: 'https://www.googleapis.com/oauth2/v4/token',
userinfo_endpoint: 'https://www.googleapis.com/oauth2/v3/userinfo',
jwks_uri: 'https://www.googleapis.com/oauth2/v3/certs',
}); // => Issuer
console.log('Set up issuer %s', googleIssuer);
Now you can create your Client.
You should provide at least the following metadata; client_id, client_secret
. You can also provide
id_token_signed_response_alg
(defaults to RS256
) and token_endpoint_auth_method
(defaults to
client_secret_basic
).
const client = new googleIssuer.Client({
client_id: 'zELcpfANLqY7Oqas',
client_secret: 'TQV5U29k1gHibH5bx1layBo0OSAvAbRT3UYW3EWrSYBB5swxjVfWUa1BS8lqzxG/0v9wruMcrGadany3'
}, [keystore]); // => Client
keystore
is an optional argument for instantiating a client with configured asymmetrical
ID Token or UserInfo response encryption.
Should your oidc provider have provided you with a registration client uri and registration access token you can also have the Client discovered.
googleIssuer.Client.fromUri(registration_client_uri, registration_access_token, [keystore]) // => Promise
.then(function (client) {
console.log('Discovered client %s', client);
});
keystore
is an optional argument for instantiating a client through registration client uri
with configured asymmetrical ID Token or UserInfo response encryption.
client.authorizationUrl({
redirect_uri: 'https://client.example.com/callback',
scope: 'openid email',
}); // => String (URL)
You can also get HTML body of a self-submitting form to utilize POST to the authorization url with
#authorizationPost
method, same signature as #authorizationUrl
.
client.authorizationPost({
redirect_uri: 'https://client.example.com/callback',
scope: 'openid email',
}); // => String (Valid HTML body)
client.authorizationCallback('https://client.example.com/callback', request.query) // => Promise
.then(function (tokenSet) {
console.log('received and validated tokens %j', tokenSet);
console.log('validated id_token claims %j', tokenSet.claims);
});
const state = session.state;
const nonce = session.nonce;
client.authorizationCallback('https://client.example.com/callback', request.query, { state, nonce, max_age }) // => Promise
.then(function (tokenSet) {
console.log('received and validated tokens %j', tokenSet);
console.log('validated id_token claims %j', tokenSet.claims);
});
When handling multiple response modes with one single pass you can use #callbackParams
to get the params object from the koa/express/node request object or a url string.
(http.IncomingMessage). If form_post is your response_type you need to include a body parser prior.
client.callbackParams('https://client.example.com/cb?code=code'); // => { code: 'code' };
client.callbackParams('/cb?code=code'); // => { code: 'code' };
// example koa v2.x w/ koa-body
app.use(bodyParser({ patchNode: true }));
app.use(async function (ctx, next) {
const params = client.callbackParams(ctx.request.req); // => parsed url query, url fragment or body object
// ...
});
// example express w/ bodyParser
app.use(bodyParser.urlencoded({ extended: false }));
app.use(function (req, res, next) {
const params = client.callbackParams(req); // => parsed url query, url fragment or body object
// ...
});
client.refresh(refreshToken) // => Promise
.then(function (tokenSet) {
console.log('refreshed and validated tokens %j', tokenSet);
console.log('refreshed id_token claims %j', tokenSet.claims);
});
Tip: accepts TokenSet as well as direct refresh token values;
client.revoke(token, [tokenTypeHint]) // => Promise
.then(function (response) {
console.log('revoked token %s', token, response);
});
client.introspect(token, [tokenTypeHint]) // => Promise
.then(function (response) {
console.log('token details %j', response);
});
client.userinfo(accessToken) // => Promise
.then(function (userinfo) {
console.log('userinfo %j', userinfo);
});
Tip: accepts TokenSet as well as direct access token values;
via POST
client.userinfo(accessToken, { verb: 'post' }); // => Promise
with extra query/body payload
client.userinfo(accessToken, { params: { fields: 'email,ids_for_business' } }); // => Promise
auth via query
client.userinfo(accessToken, { via: 'query' }); // => Promise
auth via body
client.userinfo(accessToken, { verb: 'post', via: 'body' }); // => Promise
userinfo also handles (as long as you have the proper metadata configured) responses that are:
- signed
- signed and encrypted (nested JWT)
- just encrypted
let claims = {
sub: 'userID',
_claim_names: {
credit_history: 'src1',
email: 'src2',
},
_claim_sources: {
src1: { endpoint: 'https://src1.example.com/claims', access_token: 'foobar' },
src2: { endpoint: 'https://src2.example.com/claims' },
},
};
client.fetchDistributedClaims(claims, { src2: 'bearer.for.src2' }) // => Promise
.then(function (output) {
console.log('claims %j', claims); // ! also modifies original input, does not create a copy
console.log('output %j', output);
// removes fetched names and sources and removes _claim_names and _claim_sources members if they
// are empty
});
// when rejected the error will have a property 'src' with the source name it relates to
let claims = {
sub: 'userID',
_claim_names: {
credit_history: 'src1',
email: 'src2',
},
_claim_sources: {
src1: { JWT: 'probably.a.jwt' },
src2: { JWT: 'probably.another.jwt' },
},
};
client.unpackAggregatedClaims(claims) // => Promise, autodiscovers JWT issuers, verifies signatures
.then(function (output) {
console.log('claims %j', claims); // ! also modifies original input, does not create a copy
console.log('output %j', output);
// removes fetched names and sources and removes _claim_names and _claim_sources members if they
// are empty
});
// when rejected the error will have a property 'src' with the source name it relates to
Use when the token endpoint also supports client_credentials or password grants;
client.grant({
grant_type: 'client_credentials'
}); // => Promise
client.grant({
grant_type: 'password',
username: 'johndoe',
password: 'A3ddj3w',
}); // => Promise
const opts = { keystore, initialAccessToken }; // both optional
issuer.Client.register(metadata, [opts]) // => opts optional, Promise
.then(function (client) {
console.log('Registered client %s, %j', client, client.metadata);
});
client.requestObject({ response_mode: 'form_post' }, {
sign: client.request_object_signing_alg,
encrypt: {
alg: client.request_object_encryption_alg,
enc: client.request_object_encryption_enc,
}
}).then(function (request) {
console.log('Nested signed and encrypted JWT Request Object %s', request)
});
Issuer.webfinger(userInput) // => Promise
.then(function (issuer) {
console.log('Discovered issuer %s', issuer);
});
Accepts, normalizes, discovers and validates the discovery of User Input using E-Mail, URL, acct, Hostname and Port syntaxes as described in Discovery 1.0.
Uses already discovered (cached) issuers where applicable.
authorizationCallback
and refresh
methods on a Client return TokenSet, when assigned an
expires_in
value a TokenSet calculates and assigns an expires_at
with the corresponding unix
time. It also comes with few helpers.
client.authorizationCallback(..., ...).then(function (tokenSet) {
console.log('tokenSet#expires_at', tokenSet.expires_at);
console.log('tokenSet#expires_in', tokenSet.expires_in);
setTimeout(function () {
console.log('tokenSet#expires_in', tokenSet.expires_in);
}, 2000);
console.log('tokenSet#expired()', tokenSet.expired());
console.log('tokenSet#claims', tokenSet.claims);
});
Once you have a Client instance, just pass it to the Strategy. Issuer is best discovered, Client passed properties manually or via an uri (see get-started).
Verify function is invoked with a TokenSet, userinfo only when requested, last argument is always the done function which you invoke once you found your user.
const Strategy = require('openid-client').Strategy;
const params = {
// ... any authorization request parameters go here
// client_id defaults to client.client_id
// redirect_uri defaults to client.redirect_uris[0]
// response type defaults to client.response_types[0], then 'code'
// scope defaults to 'openid'
}
const passReqToCallback = false; // optional, defaults to false, when true req is passed as a first
// argument to verify fn
const usePKCE = true; // optional, defaults to false, when true the code_challenge_method will be
// resolved from the issuer configuration, instead of true you may provide
// any of the supported values directly, i.e. "S256" (recommended) or "plain"
passport.use('oidc', new Strategy({ client, [params], [passReqToCallback], [usePKCE] }, (tokenset, userinfo, done) => {
console.log('tokenset', tokenset);
console.log('access_token', tokenset.access_token);
console.log('id_token', tokenset.id_token);
console.log('claims', tokenset.claims);
console.log('userinfo', userinfo);
User.findOne({ id: tokenset.claims.sub }, function (err, user) {
if (err) return done(err);
return done(null, user);
});
}));
// start authentication request
// options [optional], extra authentication parameters
app.get('/auth', passport.authenticate('oidc', [options]));
// authentication callback
app.get('/auth/cb', passport.authenticate('oidc', { successRedirect: '/', failureRedirect: '/login' }));
It is possible the RP or OP environment has a system clock skew, to set a clock tolerance (in seconds)
client.CLOCK_TOLERANCE = 5; // to allow a 5 second skew
Setting defaultHttpOptions
on Issuer
always merges your passed options with the default.
openid-client uses got for http requests with the following default request options
const DEFAULT_HTTP_OPTIONS = {
followRedirect: false,
headers: { 'User-Agent': `${pkg.name}/${pkg.version} (${pkg.homepage})` },
retries: 0,
timeout: 1500,
};
You can add your own headers, change the user-agent used or change the timeout setting
Issuer.defaultHttpOptions = { timeout: 2500, headers: { 'X-Your-Header': '<whatever>' } };
Confirm your httpOptions by
console.log('httpOptions %j', Issuer.defaultHttpOptions);
Because of the lightweight nature of got library the client will not use
environment-defined http(s) proxies. In order to have them used you'll need to either provide your own http request
implementation using the provided httpClient
setter or use the bundled request
one.
Custom implementation:
/*
* url {String}
* options {Object}
* options.headers {Object}
* options.body {String|Object}
* options.form {Boolean}
* options.query {Object}
* options.timeout {Number}
* options.retries {Number}
* options.followRedirect {Boolean}
*/
Issuer.httpClient = {
get(url, options) {}, // return Promise
post(url, options) {}, // return Promise
HTTPError, // used error constructor
};
Bundled (and maintained + tested) request implementation after you've added request to your package.json bundle:
npm install request@^2.0.0 --save
Issuer.useRequest();