GitLab has announced new capabilities aimed at helping enterprises manage “agent-driven” software delivery, alongside an expanded collaboration with Google Cloud that includes a managed GitLab offering delivered by certified partners.
The announcements were made at GitLab Transcend and centre on infrastructure and governance challenges GitLab said are emerging as organisations increase the use of AI agents in software development, particularly around context management, compliance controls and cost predictability.
Among the updates, GitLab introduced “Next Generation Source Code Management”, now in private beta. GitLab said the feature replaces repository cloning for agents with structured API access to server-side project intelligence and includes “minimum visibility” controls. The company said this can deliver up to 50x faster task execution per agent, use up to two times fewer tokens and generate up to 1000x less network traffic.
GitLab also announced “GitLab Orbit”, now in public beta, described as a unified context graph across the software lifecycle. According to GitLab, internal testing found agents using Orbit responded up to 11x faster and used up to 4.5x fewer tokens, while producing up to 45x fewer hallucinations.
A third capability, “Governance for Agents”, is now in private beta and adds AI auditing and control features intended to support compliance requirements. GitLab said the feature is designed to provide identity, policy and approval controls “around every agent action”, including visibility into inputs, reasoning, tool calls and high-risk or anomalous activity.
The company also introduced “GitLab Flex”, described as an annual commitment model combining platform seats, GitLab Credits and access to eligible new capabilities as they become available, with monthly reservations that can be adjusted without contract amendments.
Separately, GitLab said it will offer a managed GitLab platform on Google Cloud through GitLab-certified managed service providers, including Beyond and Digital Future. GitLab said the approach is designed to support “secure and sovereign” deployments for organisations with country-specific sovereignty and data residency requirements, while keeping code, pipeline and security data on infrastructure controlled by the customer.
As part of the expanded collaboration, GitLab said the latest versions of Google’s Gemini models, including Gemini 3.5, are now available in GitLab Duo Agent Platform. Google’s Gemma models, including Gemma 4, are now available for GitLab Duo Self-Hosted customers, aimed at self-managed and regulated environments.
“We are in the agentic engineering era, and it’s never been easier and faster to generate code. That speed brings with it a level of chaos that enterprises cannot afford,” said Manav Khurana, chief product and marketing officer at GitLab.
“AI agents are reshaping how software gets built, and the platform at the center of that shift needs to be one that enterprises can trust with their most sensitive workloads,” Khurana said in a separate statement on the Google Cloud collaboration.
GitLab said the expanded collaboration with Google Cloud builds on an April 2026 initiative that enabled customers to call Google models through GitLab Duo Agent Platform and count that usage toward existing Google Cloud commitments.
