GitLab has introduced four new products and features aimed at agent-driven software delivery: new source code management, a context graph, governance tools and commercial controls.
The announcement focuses on how companies manage software agents that write, review and ship code across large engineering environments. The new releases are intended to address technical limits in existing development workflows, rising governance demands and the difficulty of predicting costs as automated activity expands.
One of the main additions is Next Generation Source Code Management, now in private beta. It changes how software agents access repositories by replacing full repository clones with structured API access to project data on the server side.
Under this model, each agent retrieves only the information needed for a specific task, while visibility is limited to the minimum required scope. GitLab said that can cut task execution times by up to 50 times, reduce token use by as much as half and lower network traffic by up to 1000 times.
The shift reflects broader strain on traditional Git-based workflows when large numbers of software agents operate at once. In conventional setups, agents often clone entire repositories before making even narrow changes, creating bottlenecks in larger codebases or across many simultaneous tasks.
Context layer
GitLab Orbit has also entered public beta. The product is described as a context graph spanning the software lifecycle, linking code, work items, pipelines, deployments and production signals.
The aim is to give both engineers and agents access to the same source of contextual information. In large monorepos and multi-repository environments, missing context can lead agents to repeat work, consume more tokens and produce changes that teams later reverse.
Based on internal testing, GitLab said agents using Orbit responded up to 11 times faster, used up to 4.5 times fewer tokens and produced up to 45 times fewer hallucinations. Orbit also runs as a standalone data product with open APIs for use by third-party agents and external tools.
A customer quote released alongside the announcement pointed to early testing by Compare the Market.
“What GitLab Orbit gave us was something we'd been chasing for a while, the knowledge graph that backs an AI code reviewer that actually understands our codebase, not just the diff in front of it,” said Ryan Harvey, Head of AI Engineering, Compare the Market. “We tested it against retrieval-augmented generation and a few other approaches across real merge requests, and the gap was clear: better comment placement, better summaries of what actually changed. RAG, which we'd assumed would be the natural solution, ended up performing worse than no context at all. For us, that result spoke for itself.”
Audit controls
The third addition, Governance for Agents, has entered private beta. It adds auditing and control tools for AI-driven actions in software development environments.
As software agents take on more work, they can push code, alter dependencies and trigger deployments at a pace that is difficult for human reviewers to track. The product is designed to provide identity, policy, audit and approval mechanisms around each agent action, along with visibility into inputs, reasoning, tool calls and higher-risk activity across an organisation.
The release builds on earlier security agent tools in GitLab Ultimate that automate vulnerability triage and remediation. The new governance layer is aimed at helping companies maintain approval chains and compliance records when software agents act across development and operations workflows.
Commercial model
GitLab also introduced GitLab Flex, a purchasing model that combines platform seats, GitLab Credits and eligible new products under a single annual commitment. Customers can adjust monthly reservations across those categories without changing the contract, according to GitLab.
The model addresses a common procurement problem for companies adopting AI-related software tools, where spending commitments are often fixed before usage patterns become clear. By allowing monthly adjustments within an annual spend framework, GitLab is seeking to give customers more flexibility to shift between human user licences and consumption-based services.
The releases come as software groups across the industry test how far AI agents can move beyond coding assistance into broader delivery tasks. Those tasks increasingly include navigating source code, reviewing changes, triaging security issues and interacting with deployment systems, raising questions about oversight, cost control and the reliability of machine-generated output.
GitLab said it already spans development, security and operations workflows used by both people and software agents, giving it a position from which to add controls at multiple points in the software lifecycle.
“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, GitLab. “Reliability incidents, unpredictable spend, and compliance exposure in agent actions slow organisations down when they move fast without the controls they require. GitLab is the platform where enterprises already build and ship software, which means we sit at the intersection of every human and agent workflow touching code, pipelines, or production. With these new capabilities, GitLab is the agentic infrastructure that turns the speed of agentic coding into governed, auditable software delivery at enterprise scale.”