TL;DR
GitLab 19.0 extends full software lifecycle agent AI with Duo Agent Platform, adds SBOM-based dependency scanning, and supports Claude Opus 4.7 and Gemini models. The release targets the gap between faster code generation and slower delivery pipelines.
GitLab has released version 19.0the company’s first major release in a year built around a concept it calls intelligent orchestration. The point is that AI coding assistants have made coding faster, but reviews, pipelines, security scans and deployments remain manual bottlenecks. GitLab wants to close this gap.
The release expands on the GitLab Duo Agent Platform, which was generally released in January 2026. Duo agents now work through the full software lifecycle, from scheduling to security recovery, executing tasks in parallel instead of waiting for human handover at each stage.
The most significant new capability is the SBOM-based dependency scanner, now generally available. This gives Maven, Gradle, and Python projects full visibility of vulnerabilities across their entire dependency tree, including link dependencies that are not directly declared. According to Veracode’s 2025 Software Security Report, nearly 70 percent of critical security debt comes from third-party code.
GitLab Duo Developer, the platform’s AI coding assistant, gets more flexible triggering methods. Developers can now assign it to an issue, select “Create MR” or tag in any issue, or merge a query thread. The purpose is to authorize the agent take the job autonomously instead of requiring developers to context switch to a separate tool.
On the model front, GitLab 19.0 adds support for open source options including Claude Opus 4.7, Google’s Gemini models, and Devstral 2 and GLM-5.1 for self-hosted deployments. Gemini integration supports code review, vulnerability resolution, and CI/CD pipeline repair flows. Mistral AI is also available as a self-build modeling platform.
Group-level individual review guidelines are new. Previously, teams had to replicate view configurations on a project-by-project basis. Now one set of instructions can be applied to the entire group and its subgroups, which reduces installation costs for organizations managing dozens or hundreds of warehouses.
The release also introduces infrastructure changes. Valkey replaces Redis by default in the Linux package. Packaged Mattermost removed. Support for Ubuntu 20.04 has been discontinued. These are changes that require planning from self-managed customers to upgrade from version 18.
GitLab is deploying intelligent orchestration as a response to what it calls the paradox of artificial intelligence: individual developers are writing code faster than ever, but the overall speed of delivery hasn’t kept pace. of the company competitors face the same tension. GitHub recently froze new Copilot registrations after agent workflows undermined the economics of unlimited usage pricing.
GitLab’s answer to the economics question is GitLab Credits, a virtual currency with one dollar per credit that measures AI agent usage. Premium customers receive 12 credits per month per user. Perfect customers earn 24. Budget guards and spending limits introduced in version 18.11 give administrators direct control over spending.
The company recently rebuilt aligning with this strategy, realigning management levels and reorganizing R&D into around 60 autonomous teams. CEO Bill Staples called it an investment in the agency’s era. GitLab also reduced its footprint in the country by 30 percent.
The AI coding tools market will grow from $5.1 billion in 2024 to an estimated $12.8 billion in 2026. GitHub Copilot has about 37 percent of this market. GitLab’s claim is that the real value varies from generation to generation of code Organizing AI agents across the entire delivery pipeline and covering planning, coding, testing, security and deployment, a single platform has a structural advantage over point solutions.
GitLab 19.0 is now available for self-hosted instances. The company’s next big event, GitLab Transcend, will be held in London on June 10, where it plans to showcase more. Development based on artificial intelligence road map. The question for teams evaluating their options is whether one platform can better manage agents than a bunch of specialist tools.
