GitLab releases version 19.3 with new agentic AI and security features

GitLab Inc. has released GitLab 19.3, introducing several updates designed to scale agentic software development while maintaining enterprise-level control and security.

A key feature is the AI Gateway for the GitLab Duo Agent Platform, which now operates within GitLab Dedicated’s single-tenant SaaS infrastructure. This allows regulated industries to utilize agentic AI workloads while ensuring data residency and isolation rules remain consistent with their existing software development lifecycle. Users can also connect their own inference models to keep AI-processed data within established security boundaries.

Security enhancements include the GitLab Secrets Manager, now available as a paid add-on for GitLab.com users. It supports Kubernetes, Terraform, and OpenTofu, allowing credentials to be stored alongside the code and pipelines they support. Additionally, the update introduces bulk SAST (Static Application Security Testing) capabilities, including false positive detection and agentic vulnerability resolution, which provides developers with ready-to-merge fixes for high-severity risks.

For process automation, the new Flow Creator Agent allows users to create custom agentic flows using natural language, reducing the need for manual schema mapping. GitLab also introduced tools for organizations to set monthly usage limits on GitLab Credits to manage AI expenditure.