If you are looking for installation instructions, please refer to Install Kinto.

Kinto is built to be highly configurable. As a result, the related configuration can be verbose, but don’t worry, all configuration flags are listed below.

Environment variables

In order to ease deployment or testing strategies, Kinto reads settings from environment variables, in addition to .ini files.

The environment variables are exactly the same as the settings, but they are capitalised and . are replaced by _.

For example, kinto.storage_backend is read from environment variable KINTO_STORAGE_BACKEND (if defined of course).

All settings are read first from the environment variables, then from application .ini, and finally from internal defaults.

Feature settings

Setting name


What does it do?



If set to true, the whole service will reject every write operation. Note that with this option, the default bucket cannot be used and request will be rejected with a 405 Method Not Allowed error response.



The maximum number of requests that can be sent to the batch endpoint.



The maximum number of items to include on a response before enabling pagination. If set to None, no pagination will be used. It is recommended to set-up pagination if the server is under high load. If not defined, a collection response cannot contain more elements than defined by the kinto.storage_max_fetch_size setting.



The Python dotted location of the generator class that should be used to generate identifiers on a POST endpoint. Object type is one of bucket, collection, group, record. See How to define and use a custom ID generator?.



Experimental: Allow definition of JSON schema at the collection level, in order to validate submitted records. It is marked as experimental because the API might be subject to changes.



Experimental: Add a new /permissions endpoint to let the user grab the list of objects (buckets, collections, groups, records) on which they have read or write permission. It is marked as experimental because the API might be subject to changes.



Try to redirect resources removing slash or adding it for the root URL endpoint: /v1 redirects to /v1/ and /buckets/default/ to /buckets/default. No redirections are made when turned off.



Seconds specified in cache control headers on trailing slash redirects. Set to -1 to disable, and 0 to send no-cache explicitly.



The maximum duration of each heartbeat entry, in seconds.


kinto.readonly will disable every endpoint that is not accessed with one of GET, OPTIONS, or HEAD HTTP methods. Requests will receive a 405 Method Not Allowed error response.

The cache backend will still needs read-write privileges, in order to cache OAuth authentication states and tokens for example.


Kinto relies on three types of backends: storage, cache and permission. The settings names have a different prefix for each.

For each of them, the supported services are currently PostgreSQL, Redis, and Memory. Memcached is also available as a cache backend.


Setting name


What does it do?


The Python dotted location of the storage backend to use.



The URL to use to authenticate to the storage backend. e.g. redis://localhost:6378/1 or postgresql://user:pass@database/db



The maximum number of items that can be returned by one request to the storage backend. If no pagination is enabled, this is the maximum number of items that can be stored in a collection (otherwise some of them won’t be returned). With pagination enabled, this limitation doesn’t apply.



The size of the pool of connections to use for the storage backend.



Number of connections that can be opened beyond pool size.



Recycle connections after the given number of seconds has passed.



Number of seconds to wait before giving up on getting a connection from the pool.



Number of threads that can be in the queue waiting for a connection.

kinto.storage_backend =
kinto.storage_url = postgresql://postgres:postgres@localhost/postgres

# Safety limit while fetching from storage
# kinto.storage_max_fetch_size = 10000

# Control number of pooled connections
# kinto.storage_pool_size = 50


Setting name


What does it do?



The Python dotted location of the cache backend to use.



The URL to use to authenticate to the cache backend. e.g. redis://localhost:6378/1 or postgresql://user:pass@database/db



A prefix added to each key. Useful when having multiple Kinto using the same cache database.



The maximum size the memory cache backend will allow per process. (in bytes)



The size of the pool of connections to use for the cache backend.



Number of connections that can be opened beyond pool size.



Recycle connections after the given number of seconds has passed.



Number of seconds to wait before giving up on getting a connection from the pool.



Number of threads that can be in the queue waiting for a connection.



The space separated list of Memcached hosts.

For PostgreSQL

kinto.cache_backend = kinto.core.cache.postgresql
kinto.cache_url = postgresql://postgres:postgres@localhost/postgres

# Control number of pooled connections
# kinto.cache_pool_size = 50

For Memcached

You would need to install the memcached dependencies: pip install kinto[memcached]

kinto.cache_backend = kinto.core.cache.memcached
kinto.cache_hosts =


Setting name


What does it do?



The Python dotted location of the permission backend to use.



The URL to use to authenticate to the permission backend. e.g. redis://localhost:6379/1



The size of the pool of connections to use for the permission backend.



Number of connections that can be opened beyond pool size.



Recycle connections after the given number of seconds has passed.



Number of seconds to wait before giving up on getting a connection from the pool.



Number of threads that can be in the queue waiting for a connection.

kinto.permission_backend = kinto.core.permission.postgresql
kinto.permission_url = postgresql://postgres:postgres@localhost/postgres

# Control number of pooled connections
# kinto.permission_pool_size = 50

Bypass permissions with configuration

Permissions are usually retrieved from the permission backend. However, it is also possible to configure them from settings, and it will bypass the permission backend.

For example, for a resource named “bucket”, the following setting will enable authenticated people to create bucket records:

kinto.bucket_create_principals = system.Authenticated

The format of these permission settings is <resource_name>_<permission>_principals = comma,separated,principals.

If you want to give all permissions to a user you can use the following configuration:

# Kinto Admin Users Configuration
## Accounts
kinto.account_create_principals = account:admin
kinto.account_write_principals = account:admin
kinto.account_read_principals = account:admin

## Buckets
kinto.bucket_create_principals = account:admin
kinto.bucket_write_principals = account:admin
kinto.bucket_read_principals = account:admin

## Collections
kinto.collection_create_principals = account:admin
kinto.collection_write_principals = account:admin
kinto.collection_read_principals = account:admin

## Groups
kinto.group_create_principals = account:admin
kinto.group_write_principals = account:admin
kinto.group_read_principals = account:admin

## Records
kinto.record_create_principals = account:admin
kinto.record_write_principals = account:admin
kinto.record_read_principals = account:admin

Scheme, host, and port

By default, Kinto relies on WSGI for underlying details like host, port, or request scheme. Tuning these settings may be necessary when the application runs behind proxies or load balancers, but most implementations (such as uWSGI) provide adequate values automatically.

That said, if ever these items need to be controlled at the application layer, the following settings are available:

Check the behaviour of the server with the url value returned in the hello view.

Setting name


What does it do?



The HTTP Host used by Kinto to refer to itself. If set to None, the HTTP host is read from HTTP headers or WSGI environment.



The HTTP scheme used by Kinto to refer to itself. If set to None, the HTTP scheme is read from the HTTP headers or WSGI environment.

# kinto.http_scheme = https
# kinto.http_host =

Logging and Monitoring

Setting name


What does it do?



The Python dotted location of the StatsD module that should be used for monitoring. Useful to plug custom implementations like Datadog™.



The prefix to use when sending data to statsd.



The fully qualified URL to use to connect to the statsd host. e.g. udp://localhost:8125

Standard Logging

With the following configuration, all logs are redirected to standard output (See 12factor app):

keys = root

keys = console

keys = generic

level = DEBUG
handlers = console

class = StreamHandler
args = (sys.stdout,)
level = NOTSET
formatter = generic

format = %(asctime)s,%(msecs)03d %(levelname)-5.5s [%(name)s] %(message)s
datefmt = %H:%M:%S

Example output:

16:18:57,179 INFO  [root] Running kinto 6.1.0.dev0.
16:19:00,729 INFO  [request.summary]
16:19:22,232 WARNI [kinto.core.authorization] Permission not granted.
16:19:22,238 INFO  [request.summary]

Colored Logging

keys = color

class = logging_color_formatter.ColorFormatter

Example output:


JSON Logging

Using a JSON logging formatter, like this one, it is possible to output logs as JSON:

keys = json

class = kinto.core.JsonLogFormatter

Example output:

{"Pid": 19240, "Type": "root", "Timestamp": 1489067815875679744, "Severity": 6, "Hostname": "pluo", "Logger": "%", "EnvVersion": "2.0", "Fields": {"message": "Running kinto 6.1.0.dev0."}}
{"Pid": 19240, "Type": "root", "Timestamp": 1489067817834153984, "Severity": 4, "Hostname": "pluo", "Logger": "%", "EnvVersion": "2.0", "Fields": {"perm": "read", "userid": "", "message": "Permission not granted.", "uri": "/buckets/123"}}

Handling exceptions with Sentry

Sentry reporting can be enabled via the following settings:

kinto.sentry_dsn =
kinto.sentry_env = stage

# Integrate logging with Sentry.
# kinto.sentry_breadcrumbs_min_level = 10 # DEBUG
# kinto.sentry_events_min_level = 30 # WARNING

Or the equivalent environment variables:



The application sends an event on startup (mainly for setup check).

Monitoring with StatsD

Requires the statsd package.

StatsD metrics can be enabled (disabled by default):

kinto.statsd_url = udp://localhost:8125
# kinto.statsd_prefix = kinto-prod

Monitoring with New Relic

Requires the newrelic package.

Setting name


What does it do?



Location of the newrelic configuration file.



The environment the server runs into

New Relic can be enabled (disabled by default):

kinto.newrelic_config = /location/of/newrelic.ini
kinto.newrelic_env = prod


It is possible to extend the default Kinto behaviors by using “plugins”.

The list of plugins to load at startup can be specified in the settings, as a list of Python modules:

kinto.includes = kinto.plugins.default_bucket

Built-in plugins

What does it do?


It allows users to sign-up and authenticate using username and password (more details).


It is a Web admin UI to manage data from a Kinto server. (more details).


It enables a personal bucket default, where collections are created implicitly (more details).


Adds an endpoint to completely remove all data from the database backend for testing/staging purposes. (more details).


It tracks every action performed on objects within a bucket (more details).


It allows to authenticate users using OpenID Connect from Google, Microsoft, Auth0, etc. (more details).


It allows to limit storage per collection size, number of records, etc. (more details).

There are many available packages in Pyramid ecosystem, and it is straightforward to build one, since the specified module must just define an includeme(config) function.

See our list of community plugins.

See also: How to write a Kinto plugin? for more in-depth informations on how to create your own plugin.

Pluggable components

Pluggable components can be substituted from configuration files, as long as the replacement follows the original component API.

kinto.logging_renderer = your_log_renderer.CustomRenderer

This is the simplest way to extend Kinto, but will be limited to its existing components (cache, storage, log renderer, …).

In order to add extra features, including external packages is the way to go!


Kinto authentication mechanism is entirely pluggable. We call them authentication policies.

It is possible to enable several authentication policies. The order matters: when multiple policies are configured, the first one in the list that succeeds is picked.

The name matters: the policy name that is picked will be used as the prefix of the user ID (eg. ldap:alice).

Setting name


What does it do?


`` ``

The list of authentication policies names that are enabled. Each policy is configured using dedicated settings as explained below.



Python dotted path the authorization policy to use for the permission mecanism.

Authentication setup

Any authentication policy can be specified through configuration. The list of names in multiauth.policies is the starting point for Kinto to read the respective parameters (multiauth.policy.{name}.* settings).

multiauth.policies = google = kinto.plugins.openid.OpenIDConnectPolicy = = = UAlL-054uyh5in4b6u8jhg5o3hnj


With the built-in accounts plugin, users can sign-up and authenticate with username and password.

A common setup would be the following:

  • Anyone can create accounts

  • A specific admin can manage them all

# Enable built-in plugin.
kinto.includes = kinto.plugins.accounts

# Enable authenticated policy.
multiauth.policies = account
multiauth.policy.account.use = kinto.plugins.accounts.AccountsPolicy

# Allow anyone to create accounts.
kinto.account_create_principals = system.Everyone

# Set the session time to live in seconds
kinto.account_cache_ttl_seconds = 30

You can use the create-user command to create an admin:

$ kinto create-user --ini /etc/kinto.ini --username admin --password ThisIsN0tASecurePassword

You can then use this account:admin in your config:

# Allow anyone to create accounts.
kinto.account_create_principals = system.Everyone
# But also allow the admin to update, delete them etc.
kinto.account_write_principals = account:admin

About account management

You can set account_create_principals if you want to limit account creation to certain users. The most common situation is when you want to have a small number of administrators, who are responsible for creating accounts for other users. In this case, you should add the administrators to both account_create_principals and account_write_principals.

kinto.account_create_principals = account:admin /buckets/bid/groups/admin
kinto.account_write_principals = account:admin /buckets/bid/groups/admin

See the API docs to create accounts, change passwords etc.

About account validation

You can enable the account validation option, which will require account IDs to be valid email addresses, to which a validation email will be sent with an activation key.

kinto.account_validation = true
# Mail configuration:
# Set the sender for the validation email.
kinto.account_validation.email_sender = ""


Both the account validation and password reset need a properly configured SMTP server. To use a debug or testing mailer you may use the mail.mailer = debug or mail.mailer = testing settings. Refer to pyramid_mailer’s configuration.

You can restrict the email addresses allowed using the account_validation.email_regexp setting, and the delay for which the activation key will be valid:

# Set the regular expression used to validate a proper email address.
kinto.account_validation.email_regexp = "^[a-zA-Z0-9_.+-]+@[a-zA-Z0-9-]+\\.[a-zA-Z0-9-.]+$"
# Set the "time to live" for the activation key stored in the cache. After that
# delay the account won't be activable anymore.
kinto.account_validation.validation_key_cache_ttl_seconds = 604800  # 7 days in seconds.

Once created, the account will need to be activated before being able to authenticate, using the validate endpoint and the activation-key sent by email.

If the user was created, an email was sent to the user with the activation key, which needs to be POSTed to the validate endpoint.

Example email:

Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Subject: activate your account
Content-Disposition: inline


It is the responsability of the administrator to tell the mail recipient how to validate the account by modifying the email body template in the settings.

This could be done by providing a link to a webapp that displays a form to the user with a call to action to validate the user, which will POST the activation key to the validate endpoint.

Or the email could explain how to copy the activation code and paste it in some settings window.

The templates for the email subject and body can be customized:

kinto.account_validation.email_subject_template = "Account activation"
kinto.account_validation.email_body_template = "Hello {id},\n you can now activate your account using the following link:\n {form-url}{activation-key}"

… and they will be String.format-ted with the content of the user, an optional additional email-context provided alongside the user object, and the activation-key (see the note in account creation for an example usage).

Whatever the means, a POST to the /accounts/(user_id)/validate/(activation_key) will validate and activate the user.

Once the account is validated, another email will be sent for confirmation, rendered using the same email-context.

kinto.account_validation.email_confirmation_subject_template = "Account active"
kinto.account_validation.email_confirmation_body_template = "Your account {id} is now active"

About password reset

When the account validation option is enabled, an additional endpoint is available at /accounts/(user id)/reset-password to require a temporary reset password by email (see the API docs).

Example email sent to the user with the temporary reset password:

Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Subject: Reset password
Content-Disposition: inline


The template used for the email subject and body can be customized using the following settings:

kinto.account_validation.email_reset_password_subject_template = "Temporary reset password for {id}"
kinto.account_validation.email_reset_password_body_template = "Hello {id},\n you can use the following temporary reset password to change your password\n{reset-password}"

Those templates will be rendered using the user record fields, an optional additional email-context provided alongside the user object, and the reset-password (see the note in resetting a forgotten password for an example usage).

It is the responsability of the administrator to tell the mail recipient how to change their password using this temporary password.

This could be done by providing a link to a webapp that displays a form to the user asking for the new password and a call to action, which will POST the new password to the accounts/(user_id) endpoint.

Using this temporary reset password, one can update the account providing the new password.

This temporary reset password will be valid for the amount of seconds set in the settings:

# Set the "time to live" for the reset password stored in the cache.
kinto.account_validation.reset_password_cache_ttl_seconds = 604800  # 7 days in seconds.

OpenID Connect

First of all, you must find an Identity Provider. Google Identity Platform for example, but it may also be Auth0, Microsoft, Yahoo, Paypal, Bitbucket, Ebay, Salesforce, … or whichever platform that publishes its discovery metadata as JSON.

The google name below was chosen arbitrarily. As stated above, it will become the user ID prefix (e.g. and will appear in the OAuth authorized redirect URL.

While setting up the Identity Provider, you might have to fill some URLs related to your Kinto instance. For example, if you run a single page app on localhost:3000 that interacts with a server on localhost:8888, you should set:

  • Authorized JavaScript origins: http://localhost:3000

  • Authorized redirect URIs (aka. callback): http://localhost:8888/v1/openid/google/token?


If you use the Kinto Admin plugin, the JavaScript origin will be the same as the server (eg. http://localhost:8888) since the Admin Web page is served by the server itself.

Based on the information obtained during this setup, configure the issuer, client_id and client_secret values in Kinto settings:

kinto.includes = kinto.plugins.openid

multiauth.policies = google = kinto.plugins.openid.OpenIDConnectPolicy = = = UAlL-054uyh5in4b6u8jhg5o3hnj

At this point, Kinto should be properly configured and able to start.

OpenID Authentication should work as described in the API docs.

Advanced settings

# User ID field name (Default: `sub`) = email
# Authorization header prefix (Default: `Bearer`) = Bearer+OIDC

# User information cache expiration (Default: 1 day)
# Access token verification will be cached during that amount of time. = 86400

# Authentication state cache duration (Default: 1 hour)
# Duration given to users to fill the login form on the Identity Provider. = 3600
# State string length (balance between collisions/security and cache size) = 32

Of course, multiple OpenID providers can be enabled on the same Kinto server:

multiauth.policies = google auth0 = kinto.plugins.openid.OpenIDConnectPolicy =
# ...

multiauth.policy.auth0.use = kinto.plugins.openid.OpenIDConnectPolicy
multiauth.policy.auth0.issuer =
# ...

Legacy Basic Auth

In the first versions of Kinto, we had a built-in basicauth policy enabled by default.

Basically it generates a unique user identifier from any username/password combination using a HMAC secret. Bucket IDs are generated using salt declared in default_bucket_hmac_secret.

Even if it was convenient to get started, we decided to get rid of it because it was very confusing. But you can enable it with the follow configuration:

multiauth.policies = basicauth

kinto.userid_hmac_secret = have-you-seen-the-new-carioca
kinto.default_bucket_hmac_secret = bucket-id-random-salt-garam
multiauth.policy.basicauth.use = kinto.core.authentication.BasicAuthAuthenticationPolicy

Other Kinto plugins

  • Kinto LDAP: Validate Basic Auth provided user login and password with an LDAP server.

  • Kinto Facebook: Authentication using Facebook OAuth2 bearer tokens.

  • Kinto Portier: Authentication using an email address.

  • Kinto Hawk: Authentication using… Hawk.

  • Kinto Fxa: Authentication using Firefox Accounts OAuth2 bearer tokens.

Custom Authentication

Using the various Pyramid authentication packages, it is possible to plug in any kind of authentication.

In the following example, internal accounts, Persona, and IP Auth are all enabled:

multiauth.policies = account pyramid_persona ipauth

multiauth.policy.account.use = kinto.plugins.account.AccountsPolicy

multiauth.policy.ipauth.use = pyramid_ipauth.IPAuthentictionPolicy
multiauth.policy.ipauth.ipaddrs = 192.168.0.*
multiauth.policy.ipauth.userid = LAN-user
multiauth.policy.ipauth.principals = trusted

Permission handling and authorisation mechanisms are specified directly via configuration. This allows for customised solutions ranging from very simple to highly complex.


Kinto relies on pyramid multiauth to initialise authentication.

See also

Check out our tutorial about implementing a custom authentication


Kinto has a notification system, and the event listeners are configured using the event_handlers setting, which takes a list of aliases.

In the example below, the Redis listener is activated and will send events data in the queue Redis list.

kinto.event_listeners = redis

kinto.event_listeners.redis.use = kinto_redis.listeners
kinto.event_listeners.redis.url = redis://localhost:6379/0
kinto.event_listeners.redis.pool_size = 5
kinto.event_listeners.redis.listname = queue


It is possible to filter events by action and/or types of object. By default actions create, update and delete are notified for every kinds of objects.

kinto.event_listeners.redis.actions = create
kinto.event_listeners.redis.resources = bucket collection


Enabling push notifications to clients consists in enabling an event listener that will be in charge of forwarding events data to remote clients.

A Kinto plugin was made using the Pusher (commercial) service. See How to setup push notifications using WebSockets?.

Cross Origin requests (CORS)

Kinto supports CORS out of the box. Use the cors_origins setting to change the list of accepted origins.

Setting name


What does it do?



This List of CORS origins to support on all endpoints. By default allow all cross origin requests.

Backoff indicators

In order to tell clients to back-off (on heavy load for instance), the following flags can be used. Read more about this at Backoff header on heavy load.

Setting name


What does it do?



The Backoff time to use. If set to None, no backoff flag is sent to the clients. If set, provides the client with a number of seconds during which it should avoid doing unnecessary requests.



If specified, then send the backoff header with probability equal to the backoff_percentage. This should be a number between 0 and 100. This setting will have no effect if the backoff is None.



The number of seconds after which the client should issue requests.

# kinto.backoff = 10
kinto.retry_after_seconds = 30

Similarly, the end of service date can be specified by using these settings.

Setting name


What does it do?



The End of Service Deprecation date. If the date specified is in the future, an alert will be sent to clients. If it’s in the past, the service will be declared as decomissionned. If set to None, no End of Service information will be sent to the client.



The End of Service message. If set to None, no End of Service message will be sent to the clients.



The End of Service information URL.

kinto.eos = 2015-01-22
kinto.eos_message = "Client is too old"
kinto.eos_url = http://website/info-shutdown.html

Enabling or disabling endpoints

Specific resource operations can be disabled.

To do so, a setting key must be defined for the disabled resources endpoints:



  • endpoint_type is either plural (e.g. /buckets) or object (e.g. /buckets/abc);

  • resource_name is the name of the resource (e.g. bucket, group, collection, record);

  • method is the http method (in lower case) (e.g. get, post, put, patch, delete).

For example, to disable the POST on the list of buckets and DELETE on single records, the following setting should be declared in the .ini file:

kinto.plural_bucket_post_enabled = false
kinto.object_record_delete_enabled = false

Activating the permissions endpoint

The Permissions endpoint is used to get a list of all user accessible objects in the server as well as their permissions. It enables applications such as the kinto-admin to discover what the user is allowed to do and which data can be managed.

kinto.experimental_permissions_endpoint = true

Then, issue a GET request to the /permissions endpoint to get the list of the user permissions on the server ressources.

Client caching

In addition to per-collection caching, it is possible to add cache control headers for the root URL and every Kinto object. The client (or cache server or proxy) will use them to cache the collection records for a certain amount of time, in seconds.

For the root URL endpoint, when the instance is readonly, the cache control header can be adjusted with this setting:

# default is 1 day
kinto.root_cache_expires_seconds = 86400

For GET requests on resource endpoints, on any kind of object (GET /buckets, GET /buckets/{}/groups, GET /buckets/{}/collections, GET /buckets/{}/collections/{}/records), the settings are:

# kinto.bucket_cache_expires_seconds = 3600
# kinto.group_cache_expires_seconds = 3600
# kinto.collection_cache_expires_seconds = 3600
kinto.record_cache_expires_seconds = 3600

Cache can also be specified for the records of a specific bucket or collection: = 30 = 3600

If set to 0 then the resource becomes uncacheable (no-cache).


In production, Nginx can act as a cache-server using those client cache control headers.

Project information

Setting name


What does it do?



Location of the file containing the information to be shown in the version endpoint.


The HTTP link returned when uncaught errors are triggered on the server.


The URL where the documentation of the Kinto instance can be found. Will be returned in the hello view.



The name of your project (powered by Kinto)



The version of the project. Will be returned in the hello view. By default, this is the major version of Kinto.



By default, all endpoints exposed by Kinto are prefixed by a version number. If this flag is enabled, the server will redirect all requests not matching the supported version to the supported one.



Seconds specified in cache control headers on version prefix redirects. Set to -1 to disable, and 0 to send no-cache explicitly.


kinto.project_docs =
# kinto.project_version = 1.0

Application profiling

It is possible to profile the stack while its running. This is especially useful when trying to find bottlenecks.

Update the configuration file with the following values:

kinto.profiler_enabled = true
kinto.profiler_dir = /tmp/profiling

Run some request on the server (for example):

http GET http://localhost:8888/v1/

Render execution graphs using GraphViz:

sudo apt-get install graphviz
pip install gprof2dot
gprof2dot -f pstats | dot -Tpng -o output.png