GitLab 19.1: AI-воркфлоу запускаются автоматически по событиям без участия человека

GitLab 19.1 transitions AI-workflows from a mode of "someone clicks a button" to a mode of "it runs on its own"—while simultaneously giving enterprise teams tools to control what exactly is running.

Four triggers for Duo Flows

Until 19.1, each Duo Flow required manual action in the UI: mention, assign, or add a reviewer. Teams running production workflows at scale faced specific limitations: no automatic summaries of conflicts, no compliance checks on MRs ready for review, no incident creation when pipelines fail—all of this was unavailable without a human in the chain. Now GitLab 19.1 adds four event-based triggers:

• Merge request conflict — the flow launches at the moment a conflict is detected, when automatic summaries and resolution suggestions are most useful—before the developer has even opened the MR.
• Draft transitions to "ready for review" — the trigger fires when a developer removes draft status, automatically launching a compliance requirements check.
• MR approved — post-checks run automatically: deployment readiness, compliance logging, handoff notifications.
• New work item created — immediate sorting, automatic label assignment, and routing without external webhook workarounds.

The first two triggers are activated by default: teams get value immediately after upgrading to 19.1. Pipeline state filtering has also become more precise: the trigger can be configured to fire only on failure, only on success, or only on cancellation—instead of reacting to every state change.

Agent management in the environment

For regulated organizations, a non-compliant agent is not a trifle but a blocker. A single custom flow that appears in a namespace during a hackathon without security review can jeopardize the entire enterprise trial. In 19.1, two new administrative parameters appeared:

• Disable custom agents and flows — prevents users from creating or enabling custom agents and flows, restricting them to the platform's base content.
• Restrict the AI catalog to your group hierarchy — blocks activation of AI Catalog elements from other namespaces, including community and third-party content.

Together, these settings bring AI agents under the same administrative control that already applies to other sensitive platform capabilities. Widespread AI adoption becomes possible without the risk of uncontrolled agent proliferation in production.

Validation before saving

Event-based flows that launch automatically raise the cost of configuration errors. An improperly configured flow creates noise and incidents across the organization—GitLab 19.1 moves the check to the save point. When saving or updating a flow, the platform validates the configuration against the Duo Workflow Service. If something is wrong—missing input, unknown tool parameter—structured errors appear directly in the UI. The problem is detected at save time, not at 3 AM when a pipeline fails.

"Every flow in the AI

Catalog is correctly configured before anyone starts relying on it in production," — from the official GitLab 19.1 release.

Separately, a list of permitted AI models entered public beta. Administrators limit the set of providers—for example, by data residency requirements or internal security policies—and set a default model for the entire organization, while preserving user flexibility in choosing within the allowed list.

What this means

GitLab is systematically closing the main barrier to enterprise AI adoption: the inability to answer security services the question "what exactly is running in our environment and who deployed it." Automatic triggers, administrative control, and configuration validation turn AI-workflows from tools into infrastructure—predictable, manageable, and production-ready.