GitLab Announces New Features to Support Agent-Powered Software Delivery at Enterprise Scale

GitLab (headquartered in San Francisco, CA, NASDAQ: GTLB), the most comprehensive and intelligent enterprise DevSecOps platform enabling software innovation, announced new capabilities at its 'GitLab Transcend' event in London on June 10 (UK time). These features provide the infrastructure, context, and control engineering teams need to operate agent-led software delivery at enterprise scale. The Japanese version of GitLab Transcend will be held on June 18.

As engineering teams scale their use of agents, infrastructure, governance, and commercial models built for human pace are reaching their limits. The four newly announced features address bottlenecks, lack of context, compliance risks, and unpredictable costs that arise at agent scale.

Next-Generation Source Code Management (Private Beta): Up to 50x faster task execution for agents

Git was designed with human-paced operations in mind. When agents clone entire repositories to read or modify files, bottlenecks emerge as scale increases. As a result, some AI labs have built custom systems around existing Git providers to keep workflows running.

With Next-Generation Source Code Management (Private Beta), agents can query repositories server-side for only the information needed per task, with each agent’s permissions restricted to the minimum read access required for its assigned task. This increases task completion speed by up to 50x, reduces token consumption by up to 50%, and cuts network traffic by up to 1,000x.

GitLab Orbit (Public Beta): Delivers an integrated context graph

In large monorepos or multi-repo environments, agents without full lifecycle context may engage in excessive iteration, consume tokens reconstructing inaccessible information, and fill their context windows, leading to failed operations. In some cases, the time teams spend reviewing and correcting agent-generated changes exceeds the time saved by the agent.

GitLab Orbit (Public Beta) maps code, work items, pipelines, deployments, and production signals to build a context graph across the software lifecycle. This enables both agents and engineers to reference a single, trusted source of truth. Internal testing showed that using GitLab Orbit increased agent response speed by up to 11x, reduced token consumption by up to 4.5x, and decreased hallucinations by up to 45x. As a standalone data product with open APIs, GitLab Orbit allows third-party agents and external tools to leverage the same context layer.

Agent Governance (Private Beta): Auditable agent actions

The speed of agent operations exceeds existing controls. Hundreds of agents simultaneously pushing code, changing dependencies, and triggering deployments make it difficult for teams to maintain governance. This pace risks breaking the chain of management and auditability—making it hard to track which agent acted under which policy and who approved it, or to consistently define when agents can operate autonomously, when human review is required, or when activity should be halted.

GitLab recently added a Security Agent to GitLab Ultimate to automate vulnerability triage and remediation. Agent Governance (Private Beta) extends this foundation, delivering new AI audit and control capabilities to meet compliance requirements. This feature embeds identity, policy, audit, and approval mechanisms for all agent actions, enabling real-time visibility into organizational inputs, inferences, tool calls, and high-risk or anomalous activities.

GitLab Flex: Consolidates seats and credits into a single annual commitment

Many enterprise software contracts require teams to commit spending before understanding actual usage.

GitLab Flex (now available for order) covers platform seats, GitLab credits, and future eligible features under a single annual commitment. Organizations can adjust three monthly reservation tiers based on changing needs without modifying the contract.

More details on this announcement can be found here.

Executive Comments

Ryan Harvey, Head of AI Engineering at Compare the Market:

"GitLab Orbit delivered the knowledge graph we’ve long sought—one that supports AI code reviewers capable of understanding not just diffs but the entire codebase. We ran comparative tests against RAG (Retrieval-Augmented Generation) and other approaches on real merge requests, and the difference was stark. Comment placement was accurate, and summaries of actual changes were superior. RAG, which we thought was optimal, performed worse than no context at all. For us, that result clearly demonstrated GitLab Orbit’s superiority."

Manav Khurana, Chief Product and Marketing Officer (CPMO) at GitLab:

"We are now in the era of agent-driven engineering. Code generation is easier and faster than ever before. But that speed can introduce unacceptable chaos for enterprises. Responding to reliability incidents, unpredictable spending, and compliance risks from agent actions become shackles for organizations that prioritize speed over control. GitLab is a proven platform already used by organizations to build and deliver software, where workflows for everyone and every agent—across code, pipelines, and production—converge. With these new features, GitLab becomes an agent infrastructure that transforms the speed of agent coding into controlled, auditable, enterprise-scale software delivery."