GitLab has released version 19.0 of its DevSecOps platform. The new release focuses primarily on further integrating AI agents throughout the entire software development process. With this, the company aims not only to accelerate code generation but also to further automate the manual steps involved in reviews, security checks, and deployments.
According to GitLab, a new problem is emerging within many development teams: thanks to AI, developers are producing code faster, while validation, testing, and deployment continue to take a relatively long time. With version 19.0, the company is attempting to reduce that delay by deploying AI agents at multiple points within the software chain, reports TNW.
Central to this is the GitLab Duo Agent Platform, which became generally available earlier this year. In the new version, the agents can perform tasks during various phases of software development, from planning and code generation to security remediation. This means processes should run in parallel more often rather than through sequential manual hand-offs between teams.
GitLab Duo Developer, the AI assistant for developers, is also gaining new capabilities. Users can now directly link the assistant to issues or merge requests. Additionally, the agent can be invoked within discussion threads, after which it can independently generate changes or proposals.
SBOM scanning aims to improve security
One of the most significant additions is the general availability of dependency scanning based on SBOMs, or Software Bills of Materials. This allows organizations to detect vulnerabilities within entire dependency structures, including transitive dependencies that are not explicitly included in projects.
The functionality supports Maven, Gradle, and Python environments, among others. According to GitLab, third-party code poses an increasingly significant security risk in modern software development, partly because development teams rely heavily on open-source components.
GitLab is also further expanding support for AI models. Version 19.0 now supports, among others, Anthropic’s Claude Opus 4.7 and Google’s Gemini models. Open-source alternatives such as Devstral 2 and GLM-5.1 can also be deployed within self-hosted environments.
The Gemini integration is used for code reviews, restoring CI/CD pipelines, and resolving security issues, among other things. GitLab also supports self-hosted implementations of Mistral AI models.
For larger organizations, GitLab is introducing organization-wide review policies. This means settings no longer need to be configured separately for each repository. This should reduce administrative overhead, particularly for organizations with a large number of projects.
The release also includes technical changes that may impact self-managed installations. For example, Valkey replaces Redis as the default in Linux packages, the included Mattermost installation is removed, and support for Ubuntu 20.04 ends.