Hanko backend provides an HTTP API to build a modern login and registration experience for your users. Its core features are an API for passkeys (WebAuthn), passwords, and passcodes, as well as JWT management.

Hanko backend can be used on its own or in combination with hanko-elements, a powerful frontend library that contains polished and customizable UI flows for password-based and passwordless user authentication that can be easily integrated into any web app with as little as two lines of code.

API features
Running the backend
Running tests
Additional topics
API specification
Configuration reference
License

API features

Passkeys (WebAuthn)
Passcodes
Passwords
Email verification
JWT management
User management
3rd-party identity providers
Webhooks
SAML

Upcoming features

2FA configurations (optional, mandatory)

Running the backend

Note

If you just want to jump right into the experience of passkeys and passcodes, head over to the quickstart guide.

To get the Hanko backend up and running you need to:

Run a database
Configure database access
Apply database migrations
Run and configure an SMTP server
Configure JSON Web Key Set generation
Configure WebAuthn
Configure CORS
Start the backend

Run a database

The following databases are currently supported:

PostgreSQL
MySQL

Postgres

Use Docker to run a container based on the official Postgres image:

docker run --name=postgres \
-e POSTGRES_USER=<DB_USER> \
-e POSTGRES_PASSWORD=<DB_PASSWORD> \
-e POSTGRES_DB=<DB_DATABASE> \
-p <DB_PORT>:5432 \
-d postgres

or use the official binary packages to install and run a Postgres instance.

MySQL

Use Docker to run a container based on the official MySQL image:

docker run --name=mysql \
-e MYSQL_USER=<DB_USER> \
-e MYSQL_PASSWORD=<DB_PASSWORD> \
-e MYSQL_DATABASE=<DB_DATABASE> \
-e MYSQL_RANDOM_ROOT_PASSWORD=true \
-p <DB_PORT>:3306 \
-d mysql:latest

or follow the official installation instructions to install and run a MySQL instance.

Configure database access

Open the config.yaml file in the backend/config or create your own *.yaml file and add the following:

database:
  user: <DB_USER>
  password: <DB_PASSWORD>
  host: localhost # change this if the DB is not running on localhost, esp. in a production setting
  port: <DB_PORT>
  database: <DB_DATABASE>
  dialect: <DB_DIALECT> # depending on your choice of DB: postgres, mysql

Replace <DB_USER>, <DB_PASSWORD>, <DB_PORT>, <DB_DATABASE> with the values used in your running DB instance (cf. the Docker commands above used for running the DB containers) and replace <DB_DIALECT> with the DB of your choice.

Apply Database migrations

Before you can start and use the service you need to run the database migrations:

Docker

docker run --mount type=bind,source=<PATH-TO-CONFIG-FILE>,target=/config/config.yaml -p 8000:8000 -it ghcr.io/teamhanko/hanko:latest migrate up

Note

The <PATH-TO-CONFIG-FILE> must be an absolute path to your config file created above.

From source

First build the Hanko backend. The only prerequisite is to have Go (v1.18+) installed on your computer.

go generate ./...
go build -a -o hanko main.go

This command will create an executable with the name hanko, which then can be used to apply the database migrations and start the Hanko backend.

To apply the migrations, run:

./hanko migrate up --config <PATH-TO-CONFIG-FILE>

Note

The path to the config file can be relative or absolute.

Run and configure an SMTP server

The Hanko backend requires an SMTP server to send out mails containing passcodes (e.g. for the purpose of email verification, password recovery).

For local development purposes you can use, e.g., Mailslurper. Follow the official installation instructions or use an (inofficial) Docker image to get it up and running:

docker run --name=mailslurper -it -p 2500:2500 -p 8080:8080 -p 8085:8085 @marcopas/docker-mailslurper

where in this case

2500 is the SMTP port of the service
8080 is the port for the GUI application for managing mails
8085 is the port for the API service for managing mails

When using the above Docker command to run a Mailslurper container, it does not configure a user/password, so a minimal configuration in your configuration file (backend/config/config.yaml or your own *.yaml file) could contain the following:

email_delivery:
  enabled: true
  email:
    from_address: no-reply@example.com
    from_name: Example Application
  smtp:
    host: localhost
    port: 2500

To ensure that passcode emails also contain a proper subject header, configure a service name:

service:
  name: Example Authentication Service

In a production setting you would rather use a self-hosted SMTP server or a managed service like AWS SES. In that case you need to supply the email_delivery.smtp.host, email_delivery.smtp.port as well as the email_delivery.smtp.user, email_delivery.smtp.password settings according to your server/service settings.

Configure JSON Web Key Set generation

The API uses JSON Web Tokens (JWTs) for authentication. JWTs are verified using JSON Web Keys (JWK). JWKs are created internally by setting secrets.keys options in the configuration file (backend/config/config.yaml or your own *.yaml file):

secrets:
  keys:
    - <CHANGE-ME>

Note

at least one secrets.keys entry must be provided and each entry must be a random generated string at least 16 characters long.

Keys secrets are used to en- and decrypt the JWKs which get used to sign the JWTs. For every key a JWK is generated, encrypted with the key and persisted in the database.

The Hanko backend API publishes public cryptographic keys as a JWK set through the .well-known/jwks.json endpoint to enable clients to verify token signatures.

Configure WebAuthn

Passkeys are based on the Web Authentication API. In order to create and login with passkeys, the Hanko backend must be provided information about the WebAuthn Relying Party.

For most use cases, you just need the domain of your web application that uses the Hanko backend. Set webauthn.relying_party.id to the domain and set webauthn.relying_party.origin to the domain including the protocol.

Important

: If you are hosting your web application on a non-standard HTTP port (i.e. 80) you also have to include this in the origin setting.

Local development example

When developing locally, the Hanko backend defaults to:

webauthn:
  relying_party:
    id: "localhost"
    display_name: "Hanko Authentication Service"
    origins:
      - "http://localhost"

so no further configuration changes need to be made to your configuration file.

Production Examples

When you have a website hosted at example.com and you want to add a login to it that will be available at https://example.com/login, the WebAuthn config would look like this:

webauthn:
  relying_party:
    id: "example.com"
    display_name: "Example Project"
    origins:
      - "https://example.com"

If the login should be available at https://login.example.com instead, then the WebAuthn config would look like this:

webauthn:
  relying_party:
    id: "login.example.com"
    display_name: "Example Project"
    origins:
      - "https://login.example.com"

Given the above scenario, you still may want to bind your users WebAuthn credentials to example.com if you plan to add other services on other subdomains later that should be able to use existing credentials. Another reason can be if you want to have the option to move your login from https://login.example.com to https://example.com/login at some point. Then the WebAuthn config would look like this:

webauthn:
  relying_party:
    id: "example.com"
    display_name: "Example Project"
    origins:
      - "https://login.example.com"

Configure CORS

Because the backend and your application(s) consuming backend API most likely have different origins, i.e. scheme (protocol), hostname (domain), and port part of the URL are different, you need to configure Cross-Origin Resource Sharing (CORS) and specify your application(s) as allowed origins:

server:
  public:
    cors:
      allow_origins:
        - https://example.com

When you include a wildcard * origin you need to set unsafe_wildcard_origin_allowed: true:

server:
  public:
    cors:
      allow_origins:
        - "*"
      unsafe_wildcard_origin_allowed: true

Wildcard * origins can lead to cross-site attacks and when you include a * wildcard origin, we want to make sure, that you understand what you are doing, hence this flag.

Note

In most cases, the allow_origins list here should contain the same entries as the webauthn.relying_party.origins list. Only when you have an Android app you will have an extra entry (android:apk-key-hash:...) in the webauthn.relying_party.origins list.

Start the backend

The Hanko backend consists of a public and an administrative API (currently providing user management endpoints). These can be started separately or in a single command.

Start the public API

Docker

docker run --mount type=bind,source=<PATH-TO-CONFIG-FILE>,target=/config/config.yaml -p 8000:8000 -it ghcr.io/teamhanko/hanko:latest serve public

Using pre-built binaries

Each GitHub release (> 0.9.0) has hanko's binary assets uploaded to it. Alternatively you can use a tool like eget to install binaries from releases on GitHub:

eget teamhanko/hanko

From source

go generate ./...
go build -a -o hanko main.go

Then run:

./hanko serve public --config <PATH-TO-CONFIG-FILE>

Note

The <PATH-TO-CONFIG-FILE> must be an absolute path to your config file created above.

8000 is the default port for the public API. It can be customized in the configuration through the server.public.address option.

The service is now available at localhost:8000.

Start the admin API

In the usage section above we only started the public API. Use the command below to start the admin API. The default port is 8001, but can be customized in the configuration through the server.admin.address option.

serve admin

Warning

The admin API must be protected by an access management system.

Start both public and admin API

Use this command to start the public and admin API together:

serve all

Running tests

You can run the unit tests by running the following command within the backend directory:

go test -v ./...

Additional topics

Enabling password authentication

Password-based authentication is disabled per default. You can activate it and set the minimum password length in your configuration file:

password:
  enabled: true
  min_password_length: 8

Cross-domain communication

JWTs used for authentication are propagated via cookie. If your application and the Hanko backend run on different domains, cookies cannot be set by the Hanko backend. In that case the backend must be configured to transmit the JWT via Header (X-Auth-Token). To do so, enable propagation of the X-Auth-Token header:

session:
  enable_auth_token_header: true

Audit logs

API operations are recorded in an audit log. By default, the audit log is enabled and logs to STDOUT:

audit_log:
  console_output:
    enabled: true
    output: "stdout"
  storage:
    enabled: false

To persist audit logs in the database, set audit_log.storage.enabled to true.

Rate Limiting

Hanko implements basic fixed-window rate limiting for the passcode/init and password/login endpoints to mitigate brute-force attacks. It uses a combination of user-id/IP to mitigate DoS attacks on user accounts. You can choose between an in-memory and a redis store.

In production systems, you may want to hide the Hanko service behind a proxy or gateway (e.g. Kong, Traefik) to provide additional network-based rate limiting.

Hanko supports OAuth-based (authorization code flow) third party provider logins. The third_party configuration option contains all relevant configuration. This includes options for setting up redirect URLs (in case of success or error on authentication with a provider) that apply to both built-in and custom providers.

Built-in providers

Built-in providers can be configured through the third_party.providers configuration option. They must be explicitly enabled (i.e. providers are disabled default). All provider configurations require provider credentials in the form of a client ID (client_id) and a client secret (secret). See the guides in the official documentation for instructions on how to obtain these:

Custom OAuth/OIDC providers

Custom providers can be configured through the third_party.custom_providers configuration option. Like built-in providers they must be explicitly enabled and require a client_id and secret, which must be obtained from the respective provider. Custom providers can use either OAuth or OIDC. OIDC providers can be configured to use OIDC Discovery by setting the use_discovery option to true. An issuer must be configured too in that case. Otherwise both OAuth and OIDC providers can manually define required endpoints (authorization_endpoint, token_endpoint, userinfo_endpoint). scopes must be explicitly defined (with openid being the minimum requirement in case of OIDC providers).

Account linking

The allow_linking configuration option for built-in and custom providers determines whether automatic account linking for this provider is activated. Note that account linking is based on e-mail addresses and OAuth providers may allow account holders to use unverified e-mail addresses or may not provide any information at all about the verification status of e-mail addresses. This poses a security risk and potentially allows bad actors to hijack existing Hanko accounts associated with the same address. It is therefore recommended to make sure you trust the provider and to also enable emails.require_verification in your configuration to ensure that only verified third party provider addresses may be used.

User metadata

Hanko allows for defining arbitrary user metadata. Metadata can be categorized into three types that differ as to how they can be accessed and modified:

Metadata type	Public API	Admin API
Private	No read or write access	Read and write access
Public	Read access	Read and write access
Unsafe	Read access and write access	Read and write access

Each metadata type supports a maximum of 3,000 characters. Metadata is stored as compact JSON (whitespace is ignored). JSON syntax characters ({, :, ", }) count toward the character limit. Multibyte UTF-8 characters (like emojis or non-Latin characters) count as 1 character each.

Private metadata

Private metadata should be used for sensitive data that should not be exposed to the client (e.g., internal flags/ids, configuration, or access control details).

Private metadata can be read through the Admin API only using the Get metadata of a user endpoint.

Private metadata can be set and modified through the Admin API only by using the Patch metadata of a user endpoint.

Public metadata

Public metadata should be used for non-sensitive information that you want accessible but not modifiable by the client (e.g., certain user roles, UI preferences, display options).

Public metadata can be read through the Public API, the Admin API and in JWT templates for customizing the session JWT:

Public API:
- Public metadata is returned in the user object in the payload on the success state in a Login and Registration flow as well as in the payload on the profile_init state in a Profile flow.
- Public metadata is returned as part of the response of the Get a user by ID endpoint.
Admin API:
- Public metadata is returned as part of the response of the Get metadata of a user endpoint.
- Public metadata is returned as part of the response of the Get a user by ID endpoint.
JWT Templates:
- Public metadata can be accessed through the User context object available on session JWT customization. See Session JWT templates for more details.

Public metadata can be set and modified through the Admin API only by using the Patch metadata of a user endpoint.

Unsafe metadata

Unsafe metadata should be used for non-sensitive, temporary or experimental data that doesn't need strong safety guarantees.

Unsafe metadata can be read through the Public API, the Admin API and in JWT templates for customizing the session JWT:

Public API:
- Unsafe metadata is returned in the user object in the payload on the success state in a Login and Registration flow as well as in the payload on the profile_init state in a Profile flow.
- Unsafe metadata is returned as part of the response of the Get a user by ID endpoint.
Admin API:
- Unsafe metadata is returned as part of the response of the Get metadata of a user endpoint.
- Unsafe metadata is returned as part of the response of the Get a user by ID endpoint.
JWT Templates:
- Unsafe metadata can be accessed through the User context object available on session JWT customization. See Session JWT templates for more details.

Unsafe metadata can be set and modified through the Public API and the Admin API:

Public API:
- Unsafe metadata can be set using the patch_metadata action in the Profile flow.
Admin API:
- Unsafe metadata can be set using the Patch metadata of a user endpoint.

User import

You can import an existing user pool into Hanko using json in the following format:

[
  {
    "user_id": "799e95f0-4cc7-4bd7-9f01-5fdc4fa26ea3",
    "emails": [
      {
        "address": "koreyrath@wolff.name",
        "is_primary": true,
        "is_verified": true
      }
    ],
    "created_at": "2023-06-07T13:42:49.369489Z",
    "updated_at": "2023-06-07T13:42:49.369489Z"
  },
  {
    "user_id": "",
    "emails": [
      {
        "address": "joshuagrimes@langworth.name",
        "is_primary": true,
        "is_verified": true
      }
    ],
    "created_at": "2023-06-07T13:42:49.369494Z",
    "updated_at": "2023-06-07T13:42:49.369494Z"
  }
]

There is a json schema file located here that you can use for validation and input suggestions. To import users run:

hanko user import -i ./path/to/import_file.json

Webhooks

Webhooks are an easy way to get informed about changes in your Hanko instance (e.g. user or email updates). To use webhooks you have to provide an endpoint on your application which can process the events. Please be aware that your endpoint need to respond with an HTTP status code 200. Else-wise the delivery of the event will not be counted as successful.

Events

When a webhook is triggered it will send you a JSON body which contains the event and a jwt. The JWT contains 2 custom claims:

data: contains the whole object for which the change was made. (e.g.: the whole user object when an email or user is changed/created/deleted)
evt: the event for which the webhook was triggered

A typical webhook event looks like:

{
  "token": "the-jwt-token-which-contains-the-data",
  "event": "name of the event"
}

To decode the webhook you can use the JWKs created in Configure JSON Web Key Set generation

Event Types

Hanko sends webhooks for the following event types:

Event	Triggers on
user	user creation, user deletion, user update, email creation, email deletion, change of primary email
user.create	user creation
user.delete	user deletion
user.update	user update, email creation, email deletion, change of primary email
user.update.email	email creation, email deletion, change of primary email
user.update.email.create	email creation
user.update.email.delete	email deletion
user.update.email.primary	change of primary email
user.update.username.create	username creation
user.update.username.delete	username deletion
user.update.username.update	change of username
email.send	an email was sent or should be sent

As you can see, events can have subevents. You are able to filter which events you want to receive by either selecting a parent event when you want to receive all subevents or selecting specific subevents.

Enabling Webhooks

You can activate webhooks by adding the following snippet to your configuration file:

webhooks:
  enabled: true
  hooks:
    - callback: <YOUR WEBHOOK ENDPOINT>
      events:
        - user

Session JWT templates

You can define custom claims that will be added to session JWTs through the session.jwt_template.claims configuration option.

These claims are processed at JWT generation time and can include static values, templated strings using Go's text/template syntax, or nested structures (maps and slices).

The template has access to user data via the .User field, which includes:

.User.UserID: The user's unique ID (string)
.User.Email: Email details (optional, with .Address, .IsPrimary, .IsVerified)
.User.Username: The user's username (string, optional)
.User.Metadata: The user's public and unsafe metadata (optional)
- .User.Metadata.Public: The user's public metadata (object)
- .User.Metadata.Unsafe: The user's unsafe metadata (object)

Accessing user metadata

.User.Metadata.Public and .User.Metadata.Unsafe can be accessed and queried using GJSON Path Syntax (try it out in the playground).

Assume that a user's public metadata consisted of the following data:

{
    "display_name": "GamerDude",
    "favorite_games": [
        {
            "name": "Legends of Valor",
            "genre": "RPG",
            "playtime_hours": 142.3
        },
        {
            "name": "Space Raiders",
            "genre": "Sci-Fi Shooter",
            "playtime_hours": 87.6
        }
    ]
}

Then you could, for example, access this data in the following ways in your templates:

display_name: '{{ .User.Metadata.Public "display_name" }}'
favorite_games: '{{ .User.Metadata.Public "favorite_games" }}'
favorite_games_with_playtime_over_100: '{{ .User.Metadata.Public "favorite_games.#(playtime_hours>100)" }}'
favorite_genres: '{{ .User.Metadata.Public "favorite_games.#.genre" }}'

Note

Ensure you use proper quoting when accessing metadata. .User.Metadata.Public and .User.Metadata.Unsafe are function calls internally and the given path argument must be a string, so it must be double quoted. If you use use double quotes for your entire claim template then the path argument must be escaped, i.e.: "{{ .User.Metadata.Public \"display_name\" }}"

Example usage in YAML configuration:

role: "user"                                           # Static value
user_email: "{{.User.Email.Address}}"                  # Templated string
is_verified: "{{.User.Email.IsVerified}}"              # Boolean from user data
metadata:                                              # Nested map
  greeting: "Hello {{.User.Username}}"
  source: '{{ .User.Metadata.Public "display_name" }}' # Data read from public metadata
  ui_theme: '{{ .User.Metadata.Unsafe "ui_theme" }}'   # Data read from unsafe metadata
scopes:                                                # Slice with templated value
    - "read"
    - "write"
    - "{{if .User.Email.IsVerified}}admin{{else}}basic{{end}}"

In this example:

role is a static string ("user").
user_email dynamically inserts the user's email address.
is_verified inserts a boolean indicating email verification status.
metadata is a nested map with a static source and a templated greeting.
scopes is a slice combining static values and a conditional template.

Notes:

Custom claims are added at the top level of the session token payload.
Claims with the following keys will be ignored because they are currently added to the JWT by default:
- sub
- iat
- exp
- aud
- iss
- email
- username
- session_id
Templates must conform to valid Go text/template syntax. Invalid templates are logged and excluded from the generated token.
Boolean strings ("true" or "false") from templates are automatically converted to actual booleans.

README.md

Hanko backend

Contents

API features

Upcoming features

Running the backend

Run a database

Postgres

MySQL

Configure database access

Apply Database migrations

Docker

From source

Run and configure an SMTP server

Configure JSON Web Key Set generation

Configure WebAuthn

Local development example

Production Examples

Configure CORS

Start the backend

Start the public API

Docker

Using pre-built binaries

From source

Start the admin API

Start both public and admin API

Running tests

Additional topics

Enabling password authentication

Cross-domain communication

Audit logs

Rate Limiting

Social connections

Built-in providers

Custom OAuth/OIDC providers

Account linking

User metadata

Private metadata

Public metadata

Unsafe metadata

User import

Webhooks

Events

Event Types

Enabling Webhooks

Session JWT templates

Accessing user metadata

API specification

Configuration reference

License