GitLab adds enterprise controls for agentic AI tools

GitLab has introduced product updates aimed at giving enterprises more control over agentic software development. The changes include new options for GitLab Dedicated customers and additions in the GitLab 19.3 release.

At the centre of the update is the ability for GitLab Dedicated customers to run GitLab Duo Agent Platform within the same single-tenant environment and region as their existing deployments. This keeps AI-related processing inside the current security boundary while allowing customers to use their own models for inference.

The company is targeting organisations with stricter regulatory, residency and data-handling requirements. These customers already use GitLab Dedicated for sensitive software delivery workloads, and the new arrangement extends that setup to agent-based AI tools.

GitLab 19.3 also adds support for Secrets Manager, Flow Creator Agent, and two security features: Bulk SAST False Positive Detection and Agentic SAST Vulnerability Resolution. The release also makes usage caps for GitLab Credits and restricted visibility for custom agents and flows at GitLab group level generally available.

Dedicated environment

GitLab's Dedicated AI Gateway is now generally available. It allows the AI Gateway for GitLab Duo Agent Platform to run inside GitLab Dedicated single-tenant SaaS infrastructure rather than outside it.

This means agentic workloads follow the same residency and isolation model as the rest of the software development lifecycle within GitLab. For regulated or residency-sensitive teams, it offers a deployment model aligned with existing audit and governance frameworks.

Customers can also connect their own models for inference, keeping AI-processed data within the same trusted boundary used for code and delivery workflows.

Secrets control

GitLab Secrets Manager is in limited availability as a paid add-on billed through GitLab Credits for GitLab.com customers. The tool stores and manages credentials used both inside and outside continuous integration pipelines.

Each CI secret is scoped to the environment, branch and protection status of the relevant job. The product also now supports Kubernetes, Terraform, OpenTofu and custom tools.

The aim is to bring credentials under the same permission model as the code and pipelines already running on the platform. This reduces the need for teams to maintain a separate access-control framework for secrets held elsewhere.

Security backlog

GitLab is also adding bulk handling for static application security testing findings. The beta features let security teams select multiple findings in the vulnerability report and receive a confidence score for each one.

For findings confirmed as risks, the system can generate a fix ready to merge. Instead of writing each remediation from scratch, developers can review and merge the proposed change.

The feature is designed to help teams work through large backlogs of open vulnerabilities in a single action rather than processing issues one by one. New critical and high-severity findings can also continue to be triaged and remediated automatically as they arrive.

Flow creation

Flow Creator Agent is now generally available. It lets users describe an automation flow in plain language and receive a complete runnable flow in response.

The tool is available through Agentic Chat as a foundational agent in GitLab Duo Agent Platform. The resulting flow is ready to register from the AI Catalog, removing the need for users to work directly with the Flow Registry schema.

Flows still operate under a scoped service account with composite identity and require the Maintainer role or higher to enable them. This keeps workflow creation tied to existing governance controls.

Usage caps for GitLab Credits are also now generally available, giving organisations a way to set a monthly ceiling on spending related to agentic AI before overage charges apply. Administrators can set a subscription-level cap and define default per-user limits or overrides through the GraphQL API.

Restricted visibility for custom agents and flows per GitLab group is also now generally available. According to GitLab, this allows those assets to be accessible to members across multiple projects within a group, alongside existing per-project and public visibility settings.

Manav Khurana, Chief Product and Marketing Officer at GitLab, said the updates are intended to extend governance into AI-related development work. "These updates extend the speed and control enterprises need deeper into the regulated and data-sensitive segment of the enterprise market," he said.

He added: "Every capability we shipped this month, from where an agent runs to which secret it can touch, extends that same control into the trusted software delivery workflows enterprises already depend on."