GitLab 19.3 adds AI governance, secrets management and automated security remediation capabilities, including support for regulated and data-sensitive environments
GitLab, a software development platform for DevSecOps, has expanded its agentic AI capabilities with updates aimed at giving enterprises more control over how AI is deployed across software delivery workflows. The changes include a generally available AI Gateway for GitLab Dedicated, a secrets management service, automated SAST remediation and a tool for creating custom workflows through natural language.
The AI Gateway for GitLab Duo Agent Platform is now generally available on GitLab Dedicated, the company’s single-tenant SaaS environment. Customers using GitLab Dedicated can run the agent platform within the same tenant and region as their existing software development workloads, connect their own models for inference and keep AI-processed data within that security boundary.
The approach is aimed at organizations with regulatory or data residency requirements that need to apply existing isolation and residency controls to AI workloads. The AI Gateway runs within GitLab Dedicated’s infrastructure, allowing agentic workloads to follow the same deployment model used for the rest of the software development lifecycle.
GitLab is also making Secrets Manager available in limited availability as a paid add-on for GitLab.com customers, billed through GitLab Credits. The service manages credentials used both inside and outside CI pipelines, with secrets scoped according to the environment, branch and protection status of the job that requires access.
The service supports Kubernetes, Terraform, OpenTofu and custom tools. By keeping credentials within the same platform used to run code and pipelines, GitLab is applying a common permission model to those resources rather than requiring teams to manage a separate access structure.
Security teams also gain new options for handling large vulnerability backlogs. Bulk SAST False Positive Detection and Agentic SAST Vulnerability Resolution, now in beta, allows multiple findings in the Vulnerability Report to be assessed in a single action. GitLab provides a confidence score for each finding and can generate a ready-to-merge fix for confirmed risks.
The capability covers SAST vulnerabilities listed in the Vulnerability Report. GitLab also continues to triage and remediate new critical and high severity findings as they arrive.
Another addition in GitLab 19.3 is Flow Creator Agent, which is generally available. The tool allows users to describe an automation in plain language and receive a complete runnable flow that can be registered through the AI Catalog, without manually mapping the Flow Registry schema.
The generated flows continue to operate under a scoped service account with composite identity. Enabling them requires the Maintainer role or higher, keeping an existing access requirement around custom automation.
The release also makes GitLab Credits usage caps generally available. Organizations can set a monthly limit for agentic AI spending before overage charges apply, with administrators able to establish subscription-level limits in the Customers Portal and configure default or individual user caps through the GraphQL API.
Custom agents and flows can now also have restricted visibility at the GitLab group level, allowing members across multiple projects within that group to access them. The option adds to existing per-project and public visibility settings.
“These updates extend the speed and control enterprises need deeper into the regulated and data-sensitive segment of the enterprise market,” said Manav Khurana, chief product and marketing officer at GitLab. “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.”
The company says more than 50 million registered users and approximately 50% of the Fortune 100 trust GitLab.