Provider
Leonardo Antonio Carta (Nema)Controller
c/o Emperon Studio GmbH, Sachsenring 51a, 50677 Köln, Deutschland
This notice supplements the Privacy Policy for Nema Crew for the GitLab module. It does not repeat what the policy says for all modules. That includes the runtimes (Section 6), pseudonymisation and field masks (Section 7), the framework for modules (Section 11), backups (Section 17.6) and your rights (Section 18).
1. Provider and controller
This module is provided and operated by Leonardo Antonio Carta (sole proprietor), c/o Emperon Studio GmbH, Sachsenring 51a, 50677 Cologne, Germany, support@nema.im. It runs as a separate service on our server in Germany. We are the controller. If you use the module in an organisation, we process the data on behalf of the organisation under the Data Processing Agreement.
GitLab is not the provider of this module. If you use gitlab.com, GitLab processes the data in your account under its own responsibility and its privacy policy. If you connect your own GitLab instance, its operator is responsible for it and its rules apply.
2. What data the module processes
Access: You connect the module with a personal access token that you create in GitLab yourself. You decide in GitLab which rights it has. The instructions recommend the api scope: read and write access to everything your GitLab account can access. The module can use everything the token allows, but only uses the functions in Section 3.
When connecting we store the token, your GitLab username and, if you use your own instance, its address. Without an address the module connects to gitlab.com.
We only receive data your token has access to:
- projects with identifier, name, path, description, visibility, default branch and link; project members (names, GitLab usernames, access level, time they were added);
- issues and merge requests with title, description, status, branches, author and assignees (names, GitLab usernames) and comments with their authors;
- branches, commits and files with commit message, time, commit author (name, email address) and the content of files when an agent reads them;
- pipelines with status, branch, times and the results of their jobs;
- your GitLab to-dos;
- user accounts, when an agent checks your account or searches for people: name, GitLab username, profile text, location, website, profile link and creation date.
Events: If you set up a webhook in GitLab to your personal Nema address, GitLab sends us the data of the events you selected there. Depending on the event, it contains title, description, comment, project and link, who triggered it (name, GitLab username, email address), assignees and reviewers (names, GitLab usernames) as well as commit authors (names, email addresses) and commit messages.
3. What for
Exclusively for the functions of Nema Crew that you use and for the tasks that you or members of your organisation give an agent. The agent acts on your behalf with the rights of your token:
- reading: show projects, project members, issues, merge requests, comments, branches, commits, files, pipelines and your to-dos, search issues, projects and people, check your account. While you type in the app, Nema Crew suggests projects and people from GitLab;
- writing: create projects (private by default), create and change issues, including setting labels and assigning people, write, change and delete comments, open and merge merge requests, create, change and delete branches and files, start pipelines;
- cloning: copy a project into the agent's workspace on our server so that it can work on the code there and push changes back.
Triggers: If you have set up a webhook, these events start an agent: pipeline failed, pipeline succeeded, merge request opened, merge request merged, new issue, new comment. We check whether an event concerns you by your GitLab username, for example if you triggered a pipeline, are an assignee or are mentioned in a comment. Your own merge requests, issues and comments do not trigger an agent. Without a webhook there are no triggers. The module does not query GitLab at regular intervals.
4. Whom the data goes to
- to the module's service on our server;
- to GitLab or the GitLab instance you entered, when an agent reads or writes there;
- as a result or event to the language model of the runtime you chose, to the extent an agent needs the data for the task (Privacy Policy Section 6). In the Nema runtime these are the service providers we specify with each request, as our processors (Sections 6.1, 15 and 16). In the Claude Code and Codex runtimes these are Anthropic or OpenAI under your own contract (Sections 6.2 to 6.4);
- to everyone who has access to the project in GitLab, when an agent writes there, such as comments, issues or code. If an agent assigns something to a person, GitLab notifies them according to its settings;
- to other members and agents in the channel in which an agent uses the data;
- to Apple's or Google's push service, to the extent the data is in the first 180 characters of a message we send you as a notification (Section 12);
- to the data centre operator on whose hardware our server runs and where our backups are kept, as our processor (Sections 17.6 and 20);
- to authorities only if we are legally obliged to.
The module uses no further service providers.
Countries outside the EU: GitLab.com is hosted in the USA; GitLab Inc. is certified under the EU-US Data Privacy Framework (Art. 45 GDPR). For your own instance this depends on where it is located. For the service providers of the runtimes, Section 16 of the Privacy Policy applies.
5. Storage and protection
- Credentials: We store the token encrypted (AES-GCM) in our database, together with the username and the instance address. The module receives it from our server with each call and does not store it itself.
- Cloned projects: When cloning, the agent uses the module's access; the clone is kept in the agent's workshop (its workspace) on our server. It remains until it is deleted; disconnecting the module does not remove it.
- SSH keys: Each agent has its own SSH key. The private part is stored encrypted in our database. If you add the public part to GitLab in the app, we register it with your token on your GitLab account or as a deploy key of a project. It stays there until you remove it in the app or delete the agent. We then also remove it at GitLab, as far as GitLab allows.
- Content: The module does not store content from GitLab. It fetches it anew with each request. What an agent makes of it, meaning responses, memories, knowledge entries, the steps of its run and the card for an event that triggered it, we store like other content (Privacy Policy Sections 4 and 5).
- Caches: Incoming events are kept in our queue until they are processed. We delete completed entries after 7 days and failed ones after 30 days. We remember the identifier of each delivery so that an event does not arrive twice.
- Transmission: The connection to gitlab.com and between your apps and our server is encrypted with TLS. For your own instance, this depends on the address you enter.
- Pseudonymisation: In organisations that have switched on pseudonymisation, the module's results and the events that trigger an agent go to the language model pseudonymised, unless the organisation has switched this off for the module. For events, the organisation can hide or replace individual fields with field masks, such as names, GitLab usernames and email addresses of those who triggered them, assignees, reviewers and commit authors. Both are off by default, and pseudonymisation has limits (Privacy Policy Section 7).
6. Retention, disconnecting, revoking
- Disconnecting: If you disconnect the module in the app, we delete the token, username and address from our database immediately. After that, no agent retrieves data from GitLab through the module any more.
- Revoking at GitLab: You revoke the token in GitLab under Preferences → Access tokens. Disconnecting in the app does not revoke it; it remains valid until you revoke it or it expires. Also delete the webhook in GitLab, otherwise GitLab keeps sending events to your Nema address. We therefore recommend doing both.
- What remains: What an agent has taken over from GitLab into messages, memories or knowledge entries remains stored after disconnecting until you delete it (Privacy Policy Section 17.1). What an agent created or changed in GitLab stays there. Disconnecting does not remove SSH keys we registered at GitLab; remove them in the app or at GitLab.
- Deleting your account: With your account we immediately delete your credentials and content (Privacy Policy Section 17.5). We also remove your agents' SSH keys at GitLab, as far as GitLab allows.
- Backups: The data, including the encrypted token, remains in our backups for up to about 6 months (Privacy Policy Section 17.6).
7. Legal basis
Art. 6(1)(b) GDPR, because the module carries out the connection you set up. For third parties' data, such as authors, assignees, project members and commit authors, Art. 6(1)(f) GDPR applies, limited to the task you give the agent. In organisations we process on their behalf (Privacy Policy Section 2). Your confirmation of this notice is not consent. To end the processing, disconnect the module.
Documents
- This notice
- Data Processing Agreement
- GitLab Privacy Policy
This notice supplements the general privacy policy for this module. In the app you confirm it before connecting.