Skip to content
Outrich Data

Source system · Engineering

Sell your company's GitLab project data for AI training

Outrich Data is a direct buyer of GitLab project data from companies with 50 or more employees. Private code, merge request discussions and CI pipeline history can often be licensed to AI labs once secrets, names and third-party code are removed. Your GitLab administrators produce the export. Outrich Data does not connect to your instance or group.

What GitLab holds that AI labs value

GitLab keeps planning, code, build and deployment in one application. A project holds the repository, merge requests with resolvable discussion threads, issues and epics, pipeline definitions, job logs, environments and deployment records. Because it is one system, the records link to each other without extra work.

AI labs value that linkage. A single change can be followed from the issue, to the merge request, to a failed pipeline, to the commit that fixed the failure, to the passing pipeline and the deployment. That is a complete trajectory of software delivery, and it is what labs need to train coding agents that must build and ship, not only write code.

Job logs are a distinct asset. A failed job log paired with the commit that made it pass is a debugging example with a verified answer. Many self-managed instances also sit inside banks, manufacturers and other firms that keep code on their own servers, so the code comes from sectors rarely seen in public repositories.

What a usable export looks like

GitLab's project export produces an archive containing the repository, wiki, issues, merge requests with their discussions, labels, milestones and other project data. GitLab documents some items as left out of that archive, including job logs and build artifacts, so pipeline history beyond what the export carries is pulled through the API. Group export handles group-level structure such as epics. Self-managed administrators also have instance backup tools. What is included depends on your version, tier and role; check GitLab's current documentation.

Job logs need more care than code. A pipeline prints whatever its scripts print, and a masked variable can still appear in a log if it was transformed before printing. Logs are scanned and scrubbed line by line, and retried jobs are de-duplicated so one failure is not counted several times.

Outrich Data does not connect to your GitLab account, and it does not ask for credentials or an access token. Your company exports the agreed records itself, removes what is out of scope and approves the packaged dataset in writing before delivery.

Records and fields in a GitLab dataset

RecordTypical fieldsWhy it matters to AI labs
RepositoryFiles, branches, commit historyPrivate code and its evolution
Merge requestsDescription, diff, approvals, stateA change with its review outcome
Discussion threadsLine note, replies, resolved flagReview comments and whether they were addressed
PipelinesStages, jobs, status, durationBuild and test results for each commit
Job logsConsole output per jobFailure evidence paired with the eventual fix
DeploymentsEnvironment, commit, time, statusWhat reached production and when

Removed before licensing

What comes out of a GitLab export first.

  • CI/CD variables, runner tokens, deploy keys and cluster credentials
  • Secrets and internal addresses printed in job logs
  • Usernames, emails and commit author details
  • Confidential issues, security merge requests and vulnerability reports
  • Customer data that appears in test output or seed files
  • Container images, packages and build artifacts
  • Issues created from outside email through a service desk address

Data title search

Rights and contract checks.

Vendor terms for business software generally treat the content a customer puts into the service as the customer organization's data. Read your own agreement with the vendor to confirm. That settles ownership between you and the vendor. It does not settle it with your employees, customers and partners, which is what these checks cover.

  • Mirrored repositories. Projects that mirror an upstream source are not your code, even though they sit in your group.
  • Ownership of contributions. Confirm that staff and contractor agreements assign code to the company.
  • Client groups. Agencies often host client projects in their own instance. Those belong to the client.
  • Shared pipeline components. Templates and components pulled from other projects or vendors carry their own licenses.
  • Scanner output. Findings produced by licensed security tools may be restricted by that tool's terms, and are excluded with the vulnerability data.

This page is general information, not legal advice. Whether your company can license a given dataset depends on your contracts, your privacy notices and the law where you and your customers are. Review it with your own counsel before signing anything.

Value

What moves the price of GitLab data.

A written offer follows the rights review. The data valuation calculator gives an illustrative range first.

FactorEffect on value
Pipeline history keptInstances that retained job logs can supply failure and fix pairs. Short log retention removes that asset.
End-to-end linkageProjects where issues, merge requests and deployments are linked are worth more than code alone.
Self-managed, never publicCode that has only ever lived on your own servers is the least likely to have been seen by any model.
Mirrors and vendored codeA group full of mirrored upstream projects has less original work than its size suggests.

Questions

Selling GitLab data: what companies ask.

Are CI logs really worth including?
Yes, where they were kept. A failing job log together with the commit that fixed it is one of the more useful records for training a coding agent. It also needs the most careful scrubbing.
We run GitLab on our own servers. Does that change anything?
It gives your administrators more export options, and it usually means the code was never exposed to the public internet. The rights checks and the removal list are the same.
How is this different from the GitHub page?
The rights questions overlap. The records differ: GitLab projects usually carry pipeline, job log and deployment history in the same system, so the dataset can cover build and release as well as code review.
Does Outrich Data need an access token for our instance?
No. Outrich Data does not hold a personal, project or group access token and does not connect to your instance. Your administrators export and clean the projects.

GitLab is a trademark of its owner. Outrich Data is not affiliated with GitLab.

Find out if your data qualifies.

Eleven questions, about five minutes, no call required. You get an answer on screen and a written follow-up.

esc