Does Ray work with GitLab, and what can it do?

Yes. Ray reads GitLab for itself, starting with issues, merge requests, their diffs, notes, and commits, and more besides. Reading changes nothing, so it never stops to ask. 7 actions write back, and every one of them stops in the thread and waits for someone to approve it first.

What will Ray never do in GitLab?

Approve a merge request. Push commits, create files, or delete a branch. Archive or delete a project or a group. Cancel a pipeline or a job. None of those are judgment calls Ray makes carefully. They were left off its list when this integration was built, so they are not in front of it at all.

How do I connect GitLab to Ray?

Ask Ray in the thread. It answers with a Connect button under its words, you sign in to GitLab once with your own account, and Ray picks the request back up in the same thread once it is connected. Everyone who wants Ray acting in GitLab connects their own account, so there is no shared login to hand around.

Who does Ray act as in GitLab, and is my data safe?

Ray acts as the person asking. Reads run as you, writes run as whoever approved them, so GitLab's own history records real people rather than one shared bot. Ray sees what you can see and nothing you cannot. Access goes when the person does, every workspace is isolated at the database layer, and nothing you connect trains a model.

What happens if Ray gets something wrong in GitLab?

Nothing lands that you did not approve. The card shows the exact thing Ray is about to write to GitLab, word for word, so a wrong one is declined before it exists rather than undone afterwards. If Ray misread the ask itself, correct it in one message and the correction sticks, because it supersedes the old version rather than keeping both.

What if GitLab is down, or the connection breaks?

Out loud, in the thread that asked. If your GitLab connection is revoked or expires, or GitLab rate limits us, Ray says so where you asked and offers to reconnect, instead of retrying quietly in the background. Work that failed never counts against your tasks and never bills you.

What does the GitLab integration cost?

Nothing extra. Every integration, including GitLab, is on every plan, free included. The free plan covers 25 tasks a month plus a one-time 50-task grant that never expires, and no card. Paid plans start at $49 a month and are priced by team size rather than per seat.

Can Ray use GitLab together with my other tools?

Yes, and it is the reason Ray is one teammate rather than a folder of connectors. A single ask can read GitLab, cross-reference Linear and Jira, and land the finished result back in the thread. One brain and one memory across all of it, so what it learned in one tool is still true in the next.

Can I turn off approvals for GitLab?

No, and that is deliberate. Every action that sends or changes anything in GitLab waits for a human every time, on every plan. It is how Ray is built rather than a setting somebody can switch off on a busy afternoon.

GitLab, without opening GitLab.

File the issue, open the merge request, and read the failing job without leaving the thread.

Add to Slack for free Browse all integrations

Free to start
No card
Every write waits for you
Your own account, not a shared login

Try asking Ray

Ask in your own words.

Nothing to configure and no builder to open. Copy one of these into a thread, swap in your own GitLab details, and Ray takes it from there.

@Ray why did the last pipeline on main fail?
@Ray file an issue for that failing test with the job log in it.
@Ray open a merge request from the fix branch into main.
@Ray what merge requests are waiting on review, and for how long?

What Ray can and cannot do in GitLab

Reading is free, because reading changes nothing. Everything that writes waits for you. The rest is not on Ray’s list.

26 reads wired, 7 actions that wait for a human.

Runs on its own

What Ray reads in GitLab

Issues, merge requests, their diffs, notes, and commits
Pipelines and jobs, down to the failing job's log
Projects, branches, labels, milestones, and members

Reading changes nothing, so Ray never stops to ask whether it may look something up.

Asks first, every time

What waits for your approval

File an issue in the right project
Update an issue, or move it to another project
Open a merge request
Update a merge request
Comment on an issue or a merge request

Ray posts a card showing exactly what it is about to do. Nothing happens until someone clicks approve, and no setting turns that off.

Not on the list

Things Ray is never handed

These are left out during curation, so they are not something Ray could do and decides against. They are not in front of it at all.

Approve a merge request
Push commits, create files, or delete a branch
Archive or delete a project or a group
Cancel a pipeline or a job

Before it acts

Exactly what Ray will ask you to approve.

Not a category of permission. These are the GitLab actions Ray can take, each one stopping in the thread with the real content before it happens.

Move GitLab issue to another project
Edit a GitLab issue comment
Update GitLab merge request
Comment on GitLab issue
Open a GitLab merge request
File a GitLab issue
Update GitLab issue

The card Ray posts

Move GitLab issue to another project

Showing Issue iid, each carrying the exact content that will land.

ApproveDecline

How Ray works with GitLab

No workflow to build, and nothing to maintain when the work changes.

01
Connect GitLab
One click, right where you asked. Ray posts a connect card, you authorise GitLab with your own account, and the work you asked for picks up where it left off.
02
Ask in plain words
In the conversation you were already having. Nothing to build, no trigger to map, no tab to open.
03
Approve, and it lands
Reading runs on its own. Anything that creates, sends or spends stops first and shows you the exact content.

and the rest

Priya

@Ray the checkout bug is back. Take a look in GitLab, and tell Dana it is being looked at.

Move GitLab issue to another project

Waiting on: your approval

ApproveDecline

Works with your stack

Ray uses GitLab together with the rest of your tools.

One ask can read GitLab, cross-reference what is in these, and land the finished result back in the thread. Not four connectors taking turns, one teammate with one memory.

What does connecting GitLab to Ray actually do?

Ray is an AI teammate that works in your team chat, on Slack today. Connect GitLab and it can do the GitLab part of a request without anyone leaving the conversation: read what it needs, draft the change, and post it for approval before anything is written.

It is not a workflow builder. There is no trigger to configure and no automation to repair when your process shifts. You describe the outcome in the thread and Ray picks the actions, including actions in your other connected tools when the request spans more than one.

Every person on the team connects their own GitLab account, so Ray can only reach what that person can already reach, and the history in GitLab shows who actually asked.

Security and permissions

You authorise GitLab on its own login screen. Your password never passes through Ray.
Writes are gated in the execution path, not by a prompt. There is no setting that grants blanket approval.
Each workspace is isolated at the database level, and that isolation is tested by trying to break it.
Disconnect GitLab in a message, or uninstall Ray and every connection it held is revoked.

How Ray handles your data