Use the GitLab REST API to automate GitLab.

You can also use a partial OpenAPI definition, to test the API directly from the GitLab user interface. Contributions are welcome.

Available API resources

For a list of the available resources and their endpoints, see API resources.

For an introduction and basic steps, see How to make GitLab API calls.


GitLab provides an SCIM API that both implements the RFC7644 protocol and provides the /Users endpoint. The base URL is /api/scim/v2/groups/:group_path/Users/.


A GraphQL API is available in GitLab.

With GraphQL, you can make an API request for only what you need, and it's versioned by default.

GraphQL co-exists with the current v4 REST API. If we have a v5 API, this should be a compatibility layer on top of GraphQL.

There were some patenting and licensing concerns with GraphQL. However, these have been resolved to our satisfaction. The reference implementations were re-licensed under MIT, and the OWF license used for the GraphQL specification.

When GraphQL is fully implemented, GitLab:

  • Can delete controller-specific endpoints.
  • Will no longer maintain two different APIs.

Compatibility guidelines

The HTTP API is versioned with a single number, which is currently 4. This number symbolizes the major version number, as described by SemVer. Because of this, backward-incompatible changes require this version number to change.

The minor version isn't explicit, which allows for a stable API endpoint. New features can be added to the API in the same version number.

New features and bug fixes are released in tandem with GitLab. Apart from incidental patch and security releases, GitLab is released on the 22nd of each month. Backward-incompatible changes (for example, endpoint and parameter removal), and removal of entire API versions are done in tandem with major GitLab releases.

All deprecations and changes between versions are in the documentation. For the changes between v3 and v4, see the v3 to v4 documentation.

Current status

Only API version v4 is available. Version v3 was removed in GitLab 11.0.

How to use the API

API requests must include both api and the API version. The API version is defined in lib/api.rb. For example, the root of the v4 API is at /api/v4.

Valid API request

If you have a GitLab instance at gitlab.example.com:

curl "https://gitlab.example.com/api/v4/projects"

The API uses JSON to serialize data. You don't need to specify .json at the end of the API URL.

API request to expose HTTP response headers

If you want to expose HTTP response headers, use the --include option:

curl --include "https://gitlab.example.com/api/v4/projects"
HTTP/2 200

This request can help you investigate an unexpected response.

API request that includes the exit code

If you want to expose the HTTP exit code, include the --fail option:

curl --fail "https://gitlab.example.com/api/v4/does-not-exist"
curl: (22) The requested URL returned error: 404

The HTTP exit code can help you diagnose the success or failure of your REST request.


Most API requests require authentication, or only return public data when authentication isn't provided. When authentication is not required, the documentation for each endpoint specifies this. For example, the /projects/:id endpoint does not require authentication.

There are several ways you can authenticate with the GitLab API:

Project access tokens are supported by:

  • Self-managed GitLab Free and higher.
  • GitLab SaaS Premium and higher.

If you are an administrator, you or your application can authenticate as a specific user. To do so, use:

If authentication information is not valid or is missing, GitLab returns an error message with a status code of 401:

  "message": "401 Unauthorized"

OAuth2 tokens

You can use an OAuth2 token to authenticate with the API by passing it in either the access_token parameter or the Authorization header.

Example of using the OAuth2 token in a parameter:

curl "https://gitlab.example.com/api/v4/projects?access_token=OAUTH-TOKEN"

Example of using the OAuth2 token in a header:

curl --header "Authorization: Bearer OAUTH-TOKEN" "https://gitlab.example.com/api/v4/projects"

Read more about GitLab as an OAuth2 provider.

Personal/project access tokens

You can use access tokens to authenticate with the API by passing it in either the private_token parameter or the PRIVATE-TOKEN header.

Example of using the personal or project access token in a parameter:

curl "https://gitlab.example.com/api/v4/projects?private_token=<your_access_token>"

Example of using the personal or project access token in a header:

curl --header "PRIVATE-TOKEN: <your_access_token>" "https://gitlab.example.com/api/v4/projects"

You can also use personal or project access tokens with OAuth-compliant headers:

curl --header "Authorization: Bearer <your_access_token>" "https://gitlab.example.com/api/v4/projects"

Session cookie

Signing in to the main GitLab application sets a _gitlab_session cookie. The API uses this cookie for authentication if it's present. Using the API to generate a new session cookie isn't supported.

The primary user of this authentication method is the web frontend of GitLab itself. The web frontend can use the API as the authenticated user to get a list of projects without explicitly passing an access token.

GitLab CI/CD job token

When a pipeline job is about to run, GitLab generates a unique token and injects it as the CI_JOB_TOKEN predefined variable.

You can use a GitLab CI/CD job token to authenticate with specific API endpoints:

The token has the same permissions to access the API as the user that triggers the pipeline. Therefore, this user must be assigned to a role that has the required privileges.

The token is valid only while the pipeline job runs. After the job finishes, you can't use the token anymore.

A job token can access a project's resources without any configuration, but it might give extra permissions that aren't necessary. There is a proposal to redesign the feature for more strategic control of the access permissions.

GitLab CI/CD job token security

To make sure that this token doesn't leak, GitLab:

  • Masks the job token in job logs.
  • Grants permissions to the job token only when the job is running.

To make sure that this token doesn't leak, you should also configure your runners to be secure. Avoid:

  • Using Docker's privileged mode if the machines are re-used.
  • Using the shell executor when jobs run on the same machine.

If you have an insecure GitLab Runner configuration, you increase the risk that someone tries to steal tokens from other jobs.

Impersonation tokens

Impersonation tokens are a type of personal access token. They can be created only by an administrator, and are used to authenticate with the API as a specific user.

Use impersonation tokens an alternative to:

  • The user's password or one of their personal access tokens.
  • The Sudo feature. The user's or administrator's password or token may not be known, or may change over time.

For more information, see the users API documentation.

Impersonation tokens are used exactly like regular personal access tokens, and can be passed in either the private_token parameter or the PRIVATE-TOKEN header.

Disable impersonation

Introduced in GitLab 11.6.

By default, impersonation is enabled. To disable impersonation:

For Omnibus installations

  1. Edit the /etc/gitlab/gitlab.rb file:

    gitlab_rails['impersonation_enabled'] = false
  2. Save the file, and then reconfigure GitLab for the changes to take effect.

To re-enable impersonation, remove this configuration, and then reconfigure GitLab.

For installations from source

  1. Edit the config/gitlab.yml file:

      impersonation_enabled: false
  2. Save the file, and then restart GitLab for the changes to take effect.

To re-enable impersonation, remove this configuration, and then restart GitLab.


All API requests support performing an API request as if you were another user, provided you're authenticated as an administrator with an OAuth or personal access token that has the sudo scope. The API requests are executed with the permissions of the impersonated user.

As an administrator, pass the sudo parameter either by using query string or a header with an ID or username (case insensitive) of the user you want to perform the operation as. If passed as a header, the header name must be Sudo.

If a non administrative access token is provided, GitLab returns an error message with a status code of 403:

  "message": "403 Forbidden - Must be admin to use sudo"

If an access token without the sudo scope is provided, an error message is be returned with a status code of 403:

  "error": "insufficient_scope",
  "error_description": "The request requires higher privileges than provided by the access token.",
  "scope": "sudo"

If the sudo user ID or username cannot be found, an error message is returned with a status code of 404:

  "message": "404 User with ID or username '123' Not Found"

Example of a valid API request and a request using cURL with sudo request, providing a username:

GET /projects?private_token=<your_access_token>&sudo=username
curl --header "PRIVATE-TOKEN: <your_access_token>" --header "Sudo: username" "https://gitlab.example.com/api/v4/projects"

Example of a valid API request and a request using cURL with sudo request, providing an ID:

GET /projects?private_token=<your_access_token>&sudo=23
curl --header "PRIVATE-TOKEN: <your_access_token>" --header "Sudo: 23" "https://gitlab.example.com/api/v4/projects"

Status codes

The API is designed to return different status codes according to context and action. This way, if a request results in an error, you can get insight into what went wrong.

The following table gives an overview of how the API functions generally behave.

Request type Description
GET Access one or more resources and return the result as JSON.
POST Return 201 Created if the resource is successfully created and return the newly created resource as JSON.
GET / PUT Return 200 OK if the resource is accessed or modified successfully. The (modified) result is returned as JSON.
DELETE Returns 204 No Content if the resource was deleted successfully.

The following table shows the possible return codes for API requests.

Return values Description
200 OK The GET, PUT or DELETE request was successful, and the resource(s) itself is returned as JSON.
204 No Content The server has successfully fulfilled the request, and there is no additional content to send in the response payload body.
201 Created The POST request was successful, and the resource is returned as JSON.
304 Not Modified The resource hasn't been modified since the last request.
400 Bad Request A required attribute of the API request is missing. For example, the title of an issue is not given.
401 Unauthorized The user isn't authenticated. A valid user token is necessary.
403 Forbidden The request isn't allowed. For example, the user isn't allowed to delete a project.
404 Not Found A resource couldn't be accessed. For example, an ID for a resource couldn't be found.
405 Method Not Allowed The request isn't supported.
409 Conflict A conflicting resource already exists. For example, creating a project with a name that already exists.
412 The request was denied. This can happen if the If-Unmodified-Since header is provided when trying to delete a resource, which was modified in between.
422 Unprocessable The entity couldn't be processed.
429 Too Many Requests The user exceeded the application rate limits.
500 Server Error While handling the request, something went wrong on the server.


GitLab supports the following pagination methods:

  • Offset-based pagination. This is the default method and is available on all endpoints.
  • Keyset-based pagination. Added to selected endpoints but being progressively rolled out.

For large collections, for performance reasons we recommend keyset pagination (when available) instead of offset pagination.

Offset-based pagination

Sometimes, the returned result spans many pages. When listing resources, you can pass the following parameters:

Parameter Description
page Page number (default: 1).
per_page Number of items to list per page (default: 20, max: 100).

In the following example, we list 50 namespaces per page:

curl --request PUT --header "PRIVATE-TOKEN: <your_access_token>" "https://gitlab.example.com/api/v4/namespaces?per_page=50"

Pagination Link header

Link headers are returned with each response. They have rel set to prev, next, first, or last and contain the relevant URL. Be sure to use these links instead of generating your own URLs.

For GitLab.com users, some pagination headers may not be returned.

In the following cURL example, we limit the output to three items per page (per_page=3) and we request the second page (page=2) of comments of the issue with ID 8 which belongs to the project with ID 9:

curl --head --header "PRIVATE-TOKEN: <your_access_token>" "https://gitlab.example.com/api/v4/projects/9/issues/8/notes?per_page=3&page=2"

The response is:

HTTP/2 200 OK
cache-control: no-cache
content-length: 1103
content-type: application/json
date: Mon, 18 Jan 2016 09:43:18 GMT
link: <https://gitlab.example.com/api/v4/projects/8/issues/8/notes?page=1&per_page=3>; rel="prev", <https://gitlab.example.com/api/v4/projects/8/issues/8/notes?page=3&per_page=3>; rel="next", <https://gitlab.example.com/api/v4/projects/8/issues/8/notes?page=1&per_page=3>; rel="first", <https://gitlab.example.com/api/v4/projects/8/issues/8/notes?page=3&per_page=3>; rel="last"
status: 200 OK
vary: Origin
x-next-page: 3
x-page: 2
x-per-page: 3
x-prev-page: 1
x-request-id: 732ad4ee-9870-4866-a199-a9db0cde3c86
x-runtime: 0.108688
x-total: 8
x-total-pages: 3

Other pagination headers

GitLab also returns the following additional pagination headers:

Header Description
x-next-page The index of the next page.
x-page The index of the current page (starting at 1).
x-per-page The number of items per page.
X-prev-page The index of the previous page.
x-total The total number of items.
x-total-pages The total number of pages.

For GitLab.com users, some pagination headers may not be returned.

Keyset-based pagination

Keyset-pagination allows for more efficient retrieval of pages and - in contrast to offset-based pagination - runtime is independent of the size of the collection.

This method is controlled by the following parameters:

Parameter Description
pagination keyset (to enable keyset pagination).
per_page Number of items to list per page (default: 20, max: 100).

In the following example, we list 50 projects per page, ordered by id ascending.

curl --request GET --header "PRIVATE-TOKEN: <your_access_token>" "https://gitlab.example.com/api/v4/projects?pagination=keyset&per_page=50&order_by=id&sort=asc"

The response header includes a link to the next page. For example:

HTTP/1.1 200 OK
Links: <https://gitlab.example.com/api/v4/projects?pagination=keyset&per_page=50&order_by=id&sort=asc&id_after=42>; rel="next"
Link: <https://gitlab.example.com/api/v4/projects?pagination=keyset&per_page=50&order_by=id&sort=asc&id_after=42>; rel="next"
Status: 200 OK

WARNING: The Links header is scheduled to be removed in GitLab 14.0 to be aligned with the W3C Link specification. The Link header was added in GitLab 13.1 and should be used instead.

The link to the next page contains an additional filter id_after=42 that excludes already-retrieved records. The type of filter depends on the order_by option used, and we may have more than one additional filter.

When the end of the collection is reached and there are no additional records to retrieve, the Link header is absent and the resulting array is empty.

We recommend using only the given link to retrieve the next page instead of building your own URL. Apart from the headers shown, we don't expose additional pagination headers.

Keyset-based pagination is supported only for selected resources and ordering options:

Resource Order
Projects order_by=id only.

Path parameters

If an endpoint has path parameters, the documentation displays them with a preceding colon.

For example:

DELETE /projects/:id/share/:group_id

The :id path parameter needs to be replaced with the project ID, and the :group_id needs to be replaced with the ID of the group. The colons : shouldn't be included.

The resulting cURL request for a project with ID 5 and a group ID of 17 is then:

curl --request DELETE --header "PRIVATE-TOKEN: <your_access_token>" "https://gitlab.example.com/api/v4/projects/5/share/17"

Path parameters that are required to be URL-encoded must be followed. If not, it doesn't match an API endpoint and responds with a 404. If there's something in front of the API (for example, Apache), ensure that it doesn't decode the URL-encoded path parameters.

Namespaced path encoding

If using namespaced API requests, make sure that the NAMESPACE/PROJECT_PATH is URL-encoded.

For example, / is represented by %2F:

GET /api/v4/projects/diaspora%2Fdiaspora

A project's path isn't necessarily the same as its name. A project's path is found in the project's URL or in the project's settings, under General > Advanced > Change path.

File path, branches, and tags name encoding

If a file path, branch or tag contains a /, make sure it is URL-encoded.

For example, / is represented by %2F:

GET /api/v4/projects/1/repository/files/src%2FREADME.md?ref=master
GET /api/v4/projects/1/branches/my%2Fbranch/commits
GET /api/v4/projects/1/repository/tags/my%2Ftag

Request Payload

API Requests can use parameters sent as query strings or as a payload body. GET requests usually send a query string, while PUT or POST requests usually send the payload body:

  • Query string:

    curl --request POST "https://gitlab/api/v4/projects?name=<example-name>&description=<example-description>"
  • Request payload (JSON):

    curl --request POST --header "Content-Type: application/json" --data '{"name":"<example-name>", "description":"<example-description"}' "https://gitlab/api/v4/projects"

URL encoded query strings have a length limitation. Requests that are too large result in a 414 Request-URI Too Large error message. This can be resolved by using a payload body instead.

Encoding API parameters of array and hash types

You can request the API with array and hash types parameters:


import_sources is a parameter of type array:

curl --request POST --header "PRIVATE-TOKEN: <your_access_token>" \
-d "import_sources[]=github" \
-d "import_sources[]=bitbucket" \


override_params is a parameter of type hash:

curl --request POST --header "PRIVATE-TOKEN: <your_access_token>" \
--form "namespace=email" \
--form "path=impapi" \
--form "file=@/path/to/somefile.txt"
--form "override_params[visibility]=private" \
--form "override_params[some_other_param]=some_value" \

Array of hashes

variables is a parameter of type array containing hash key/value pairs [{ 'key': 'UPLOAD_TO_S3', 'value': 'true' }]:

curl --globoff --request POST --header "PRIVATE-TOKEN: <your_access_token>" \

curl --request POST --header "PRIVATE-TOKEN: <your_access_token>" \
--header "Content-Type: application/json" \
--data '{ "ref": "master", "variables": [ {"key": "VAR1", "value": "hello"}, {"key": "VAR2", "value": "world"} ] }' \

id vs iid

Some resources have two similarly-named fields. For example, issues, merge requests, and project milestones. The fields are:

  • id: ID that is unique across all projects.
  • iid: Additional, internal ID (displayed in the web UI) that's unique in the scope of a single project.

If a resource has both the iid field and the id field, the iid field is usually used instead of id to fetch the resource.

For example, suppose a project with id: 42 has an issue with id: 46 and iid: 5. In this case:

  • A valid API request to retrieve the issue is GET /projects/42/issues/5.
  • An invalid API request to retrieve the issue is GET /projects/42/issues/46.

Not all resources with the iid field are fetched by iid. For guidance regarding which field to use, see the documentation for the specific resource.

Data validation and error reporting

When working with the API you may encounter validation errors, in which case the API returns an HTTP 400 error.

Such errors appear in the following cases:

  • A required attribute of the API request is missing (for example, the title of an issue isn't given).
  • An attribute did not pass the validation (for example, the user bio is too long).

When an attribute is missing, you receive something like:

HTTP/1.1 400 Bad Request
Content-Type: application/json
    "message":"400 (Bad request) \"title\" not given"

When a validation error occurs, error messages are different. They hold all details of validation errors:

HTTP/1.1 400 Bad Request
Content-Type: application/json
    "message": {
        "bio": [
            "is too long (maximum is 255 characters)"

This makes error messages more machine-readable. The format can be described as follows:

    "message": {
        "<property-name>": [
        "<embed-entity>": {
            "<property-name>": [

Unknown route

When you attempt to access an API URL that doesn't exist, you receive a 404 Not Found message.

HTTP/1.1 404 Not Found
Content-Type: application/json
    "error": "404 Not Found"

Encoding + in ISO 8601 dates

If you need to include a + in a query parameter, you may need to use %2B instead, due to a W3 recommendation that causes a + to be interpreted as a space. For example, in an ISO 8601 date, you may want to include a specific time in ISO 8601 format, such as:


The correct encoding for the query parameter would be:



There are many unofficial GitLab API Clients for most of the popular programming languages. For a complete list, visit the GitLab website.

Rate limits

For administrator documentation on rate limit settings, see Rate limits. To find the settings that are specifically used by GitLab.com, see GitLab.com-specific rate limits.

Content type

The GitLab API supports the application/json content type by default, though some API endpoints also support text/plain.

In GitLab 13.10 and later, API endpoints do not support text/plain by default, unless it's explicitly documented.