GitLab Token overview (FREE)
This document lists tokens used in GitLab, their purpose and, where applicable, security guidance.
Personal access tokens
You can create Personal access tokens to authenticate with:
- The GitLab API.
- GitLab repositories.
- The GitLab registry.
You can limit the scope and expiration date of your personal access tokens. By default, they inherit permissions from the user who created them.
OAuth2 tokens
GitLab can serve as an OAuth2 provider to allow other services to access the GitLab API on a user's behalf.
You can limit the scope and lifetime of your OAuth2 tokens.
Impersonation tokens
An Impersonation token is a special type of personal access token. It can be created only by an administrator for a specific user. Impersonation tokens can help you build applications or scripts that authenticate with the GitLab API, repositories, and the GitLab registry as a specific user.
You can limit the scope and set an expiration date for an impersonation token.
Project access tokens
Project access tokens are scoped to a project. As with Personal access tokens, you can use them to authenticate with:
- The GitLab API.
- GitLab repositories.
- The GitLab registry.
You can limit the scope and expiration date of project access tokens. When you create a project access token, GitLab creates a project bot user. Project bot users are service accounts and do not count as licensed seats.
Deploy tokens
Deploy tokens allow you to download (git clone) or push and pull packages and container registry images of a project without having a user and a password. Deploy tokens cannot be used with the GitLab API.
Deploy tokens can be managed by project maintainers and owners.
Deploy keys
Deploy keys allow read-only or read-write access to your repositories by importing an SSH public key into your GitLab instance. Deploy keys cannot be used with the GitLab API or the registry.
This is useful, for example, for cloning repositories to your Continuous Integration (CI) server. By using deploy keys, you don't have to set up a fake user account.
Project maintainers and owners can add or enable a deploy key for a project repository
Runner registration tokens
Runner registration tokens are used to register a runner with GitLab. Group or project owners or instance administrators can obtain them through the GitLab user interface. The registration token is limited to runner registration and has no further scope.
You can use the runner registration token to add runners that execute jobs in a project or group. The runner has access to the project's code, so be careful when assigning project and group-level permissions.
Runner authentication tokens (also called runner tokens)
After registration, the runner receives an authentication token, which it uses to authenticate with GitLab when picking up jobs from the job queue. The authentication token is stored locally in the runner's config.toml file.
After authentication with GitLab, the runner receives a job token, which it uses to execute the job.
In case of Docker Machine/Kubernetes/VirtualBox/Parallels/SSH executors, the execution environment has no access to the runner authentication token, because it stays on the runner machine. They have access to the job token only, which is needed to execute the job.
Malicious access to a runner's file system may expose the config.toml file and thus the authentication token, allowing an attacker to clone the runner.
CI/CD job tokens
The CI/CD job token is a short lived token only valid for the duration of a job. It gives a CI/CD job access to a limited amount of API endpoints. API authentication uses the job token, by using the authorization of the user triggering the job.
The job token is secured by its short life-time and limited scope. It could possibly be leaked if multiple jobs run on the same machine (like with the shell runner). On Docker Machine runners, configuring MaxBuilds=1 is recommended to make sure runner machines only ever run one build and are destroyed afterwards. This may impact performance, as provisioning machines takes some time.
Available scopes
This table shows available scopes per token. Scopes can be limited further on token creation.
| API access | Registry access | Repository access | |
|---|---|---|---|
| Personal access token | |||
| OAuth2 token | |||
| Impersonation token | |||
| Project access token |
|
|
|
| Group access token |
|
|
|
| Deploy token | |||
| Deploy key | |||
| Runner registration token |
|
||
| Runner authentication token |
|
||
| Job token |
|
- Limited to the one project.
- Limited to the one group.
- Runner registration and authentication token don't provide direct access to repositories, but can be used to register and authenticate a new runner that may execute jobs which do have access to the repository
- Limited to certain endpoints.
Security considerations
Access tokens should be treated like passwords and kept secure.
Adding them to URLs is a security risk. This is especially true when cloning or adding a remote, as Git then writes the URL to its .git/config file in plain text. URLs are also generally logged by proxies and application servers, which makes those credentials visible to system administrators.
Instead, API calls can be passed an access token using headers, like the Private-Token header.
Tokens can also be stored using a Git credential storage.
Tokens should not be committed to your source code. Instead, consider an approach such as using external secrets in CI.
When creating a scoped token, consider using the most limited scope possible to reduce the impact of accidentally leaking the token.
When creating a token, consider setting a token that expires when your task is complete. For example, if performing a one-off import, set the token to expire after a few hours or a day. This reduces the impact of a token that is accidentally leaked because it is useless when it expires.