> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rubixkube.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# GitLab integration

> Give RubixKube agents read and write access to your GitLab repos so they can correlate CI deployments with incidents, cite MRs in RCAs, and propose soft remediations as merge requests.

Connect GitLab so RubixKube agents can see what shipped through your pipelines, correlate CI deployments with incidents, and propose soft remediations as merge requests instead of direct cluster changes.

## What it unlocks

<CardGroup cols={2}>
  <Card title="Pipelines and MRs in context" icon="rocket">
    Agents can read recent deploys, merges, and commit history. RCAs cite pipeline runs that correlate with an incident when the evidence points that way.
  </Card>

  <Card title="Merge requests as a remediation surface" icon="code-pull-request">
    For IaC and GitOps repos, agents can open MRs with the proposed change. Your review, your merge, your pipeline.
  </Card>

  <Card title="Issue tracking" icon="list-check">
    Agents can create, read, and update issues linked to insights and RCA reports.
  </Card>

  <Card title="Evidence, not source" icon="shield-halved">
    The integration reads metadata and diffs. Source code is not ingested or stored.
  </Card>
</CardGroup>

## Connect it

<Steps titleSize="h3">
  <Step title="Open the Integrations page">
    In the RubixKube console, go to **Integrations** and select **GitLab**.
  </Step>

  <Step title="Authenticate">
    Authorise with GitLab. Both GitLab.com and self-hosted instances are supported. For self-hosted, provide your GitLab base URL.
  </Step>

  <Step title="Configure access">
    Pick the groups or projects RubixKube can see. Grant write access only to projects where you want agents to open merge requests.
  </Step>
</Steps>

Once connected, GitLab is available tenant-wide. Every agent across every environment can use it without additional per-environment setup.

## How the agents use it

* During RCA, agents walk recent pipelines and merges to the affected service as candidate causes, citing them as evidence.
* For IaC and GitOps-managed infrastructure, soft remediations are proposed as merge requests against the relevant project. See [Safety and Guardrails](/concepts/guardrails) for the full approval model.
* Chat can answer questions like "what changed in payments before this spike" by walking the pipeline and merge history.

## Build custom workflows with skills

Out of the box, the integration exposes read and MR-open capabilities to agents. To program specific workflows (auto-open an MR with a suggested config fix, weekly changelog digest, project-to-service mapping rules), build a [Skill](/concepts/skills) on top of the connection.

## Self-hosted GitLab

Self-hosted instances work the same way. Provide your instance's base URL during setup. For restricted networks, reach out to [connect@rubixkube.ai](mailto:connect@rubixkube.ai) and we will walk you through the right deployment mode.

## Troubleshooting

<AccordionGroup>
  <Accordion title="Agents cannot see a project">
    Confirm the authorised account has access to the group or project, and that the project is included in the scope you chose during setup.
  </Accordion>

  <Accordion title="MR proposals fail">
    Check that the target project is included in the write scope. Reads work with a narrower scope than writes.
  </Accordion>
</AccordionGroup>

## Related guides

<CardGroup cols={2}>
  <Card title="GitHub" icon="github" href="/integrations/github">
    Equivalent integration for GitHub-hosted repos.
  </Card>

  <Card title="Skills" icon="sparkles" href="/concepts/skills">
    How to compose GitLab tools into custom workflows.
  </Card>

  <Card title="Safety and Guardrails" icon="shield-halved" href="/concepts/guardrails">
    Why MR-based remediation is the preferred path.
  </Card>
</CardGroup>