GitLab Duo Agent Platform now runs on Vertex AI

AI agents with governance: GitLab and Google Cloud join forces.

On April 14, 2026, GitLab announced an expanded partnership with Google Cloud. Google Cloud customers can now power the GitLab Duo Agent Platform using the Vertex AI models they already use, and count that usage toward their existing Google Cloud commitments. That sounds like a commercial deal, but the technical impact is greater than it seems.

What exactly is the GitLab Duo Agent Platform?

The GitLab Duo Agent Platform is the AI layer built on top of GitLab that enables autonomous software agents to perform tasks within your CI/CD pipeline (the automated process from coding to deployment). Launched in January 2026, the platform orchestrates the entire software development lifecycle by integrating agent-driven planning, coding, review, and security remediation. Until now, this platform ran on proprietary models or required separate AI infrastructure.

AI agents in the Duo Agent Platform can now call foundation models via Vertex AI, including Gemini models, with actions governed by GitLab’s built-in compliance and audit controls. Organizations can run GitLab’s AI Gateway on Google Cloud without having to set up or manage a separate AI infrastructure. This last point is particularly relevant. You don’t need to set up a separate GPU cluster to get the AI agents up and running. GitLab acts as the control layer, while Google Cloud provides the computing power.

Governance doesn't stop just because an agent takes over the work.

For IT decision-makers, there is a crucial question that arises with every AI application in software development: who is responsible if an agent makes a mistake? When an agent generates a code suggestion, it draws context from the issue that triggered the work, the existing patterns in the repository, and the CI pipeline that validates the result—all without leaving GitLab. Every action an agent takes goes through the same access controls, approval rules, and audit logs that developers already follow, so governance doesn’t stop when an agent takes over a task.

The Vertex AI Model Garden offers joint customers a choice of multiple foundation models, including Gemini, allowing teams to select models based on performance, cost, or regulatory requirements. Consider organizations in the healthcare or financial sectors that are subject to data residency requirements: you can choose the model that fits your compliance landscape.

What does this mean for Cube and its clients?

Cube is an official GitLab Channel Partner and uses GitLab as the core of its DevSecOps approach. DevSecOps refers to the integration of security at every stage of software development, from the first commit to production. This integration enables Cube’s clients to deploy AI agents for tasks such as automated code review, vulnerability scanning, and incident response, while ensuring that all actions remain traceable and auditable within the existing GitLab environment.

For B2B organizations that handle sensitive data or operate under strict SLAs (Service Level Agreements), this offers a clear path to productive AI without compromising on control and accountability.