GitLab 19.0 aims to bridge the divide between coding and delivery.

      **TL;DR:** GitLab 19.0 broadens its agentic AI capabilities throughout the entire software lifecycle with the Duo Agent Platform, introduces SBOM-based dependency scanning, and incorporates support for Claude Opus 4.7 and Gemini models. This release aims to bridge the divide between rapid code generation and slower delivery processes.

      GitLab has launched version 19.0, marking its first major update in a year, centering around what the company refers to as intelligent orchestration. The premise suggests that while AI coding assistants accelerate code writing, the processes of reviews, pipelines, security checks, and deployments continue to be manual hurdles. GitLab aims to address this issue.

      The update enhances the GitLab Duo Agent Platform, which became generally available in January 2026. Duo agents can now operate throughout the software lifecycle, from planning to security fixes, by executing tasks concurrently instead of pausing for human intervention at each step.

      A key new feature is the SBOM-based dependency scanner, now widely available. This tool enables Maven, Gradle, and Python projects to gain complete insight into vulnerabilities throughout their entire dependency tree, including transitive dependencies that aren't directly listed. This is significant as about 70 percent of critical security liabilities arise from third-party code, based on Veracode’s 2025 State of Software Security report.

      GitLab Duo Developer, the platform's AI coding assistant, has gained more adaptable trigger options. Developers can now assign it to an issue, choose "Generate MR," or mention it in any issue or merge request discussion thread, allowing the agent to autonomously take on tasks without requiring developers to switch contexts.

      In terms of models, GitLab 19.0 includes support for Claude Opus 4.7, Google’s Gemini models, and self-hosted open-source solutions like Devstral 2 and GLM-5.1. The Gemini integration assists with code reviews, resolving vulnerabilities, and repairing CI/CD pipeline flows. Mistral AI is also now accessible as a self-hosted modeling platform.

      New group-level custom review instructions have been introduced. Previously, teams had to replicate review settings for each project, but now a single set of instructions can be applied across an entire group and its subgroups, streamlining the setup process for organizations managing numerous repositories.

      Infrastructure updates are also part of this release. Valkey is now the default in the Linux package, replacing Redis, and bundled Mattermost has been removed. Support for Ubuntu 20.04 has been discontinued. These changes will require careful planning for self-managed customers upgrading from version 18.

      GitLab positions intelligent orchestration as a solution to what it calls the AI paradox: individual developers are coding faster than ever, yet overall delivery speed has not increased correspondingly. Competitors face similar challenges; for instance, GitHub recently halted new Copilot sign-ups after finding that agentic workflows disrupted its pricing model.

      To address economic concerns, GitLab has introduced GitLab Credits, a virtual currency costing one dollar per credit to regulate AI agent usage. Premium users receive 12 credits monthly, while ultimate users get 24. Budget controls and spending limits, introduced in version 18.11, provide administrators with direct oversight of expenditures.

      The company has recently reorganized to align with this strategy, flattening management layers and restructuring research and development into approximately 60 independent teams. CEO Bill Staples described this as an investment in the agentic era. GitLab has also reduced its presence in various countries by 30 percent.

      The market for AI coding tools is projected to reach around $12.8 billion in 2026, up from $5.1 billion in 2024, with GitHub Copilot holding about 37 percent of the market share. GitLab's belief is that the true value is transitioning from code generation to orchestrating AI agents throughout the entire delivery pipeline, suggesting that a comprehensive platform covering all aspects—planning, coding, testing, security, and deployment—holds a distinct advantage over specialized tools.

      GitLab 19.0 is now available for self-managed instances. The company’s next major event, GitLab Transcend, is set to take place on June 10 in London, where it plans to reveal more about its AI-driven development strategy. Teams considering their options must evaluate whether a single platform can effectively orchestrate agents better than a collection of specialized tools.