GitLab Pitches Git-Compatible SCM for Thousands of AI Agents

Why agents strain source control

According to GitLab, AI agents are changing the economics of software development. As the cost of producing code falls, the volume of changes moving through the software factory is growing faster than infrastructure designed for human developers can handle.

The company points to Anthropic as an example. Daily merges to Anthropic's monorepo grew from approximately 2,500 in May to more than 30,000 in September, and its engineers built Scion, a specialized source control system for their environment. GitLab says Scion is purpose-built for enormous monorepos where nearly all clients are agents that are always online. Anthropic acknowledges that its approach is not a general-purpose replacement for Git and requires its own code review, search, and CI infrastructure, GitLab writes.

GitLab quotes Nick G. of Anthropic's source control team: "Scaling our monorepo for agentic development has changed how we think about Git server architecture. We see room to expand what Git servers can do for these workloads without departing from the overall Git model."

What next-gen SCM changes

GitLab argues that traditional Git workflows make clients download substantial repository data even when an agent needs only a small part. By the company's account, a single agent invocation can involve 5 to 10 GB of data transfer and more than 30 seconds of setup, just to answer one question. Across thousands of agents, source control becomes a bottleneck.

Next-gen SCM keeps the Git protocol but changes the architecture underneath. Stateless compute scales horizontally against shared object storage, and agent-native interfaces let agents request the files and history they need without repeatedly cloning entire repositories.

In early internal tests, GitLab says workloads completed up to 50x faster, used up to 2x fewer tokens, and generated up to 1,000x less network traffic compared with conventional workflows. The company stresses that these are peak improvements under specific test conditions and that it is still validating performance under sustained, concurrent load.

The architecture is designed to run on GitLab.com and on S3-compatible storage that customers control, including in air-gapped data centers. According to GitLab, repository data stays within the customer's infrastructure, which preserves data sovereignty and existing investments.

The bottleneck moves downstream

GitLab notes that faster source control alone only shifts the problem. An agent that writes code in seconds creates little value if changes wait hours for validation, security review, or deployment. Merge queues and CI workflows face growing pressure when agents generate changes faster than validation can finish.

The company says it sees this on GitLab.com: over the past year, customers created 40% more CI/CD pipelines, while code pushes increased 50%. It expects agent-scale CI to validate more changes with less compute per change while preserving evidence of what was tested and verified.

GitLab describes next-gen SCM as one part of a broader vision it calls a governed software factory. The design lets thousands of agents work in parallel in isolated sessions, where failed experiments can be discarded without disrupting shared code. Successful changes can flow into existing merge requests, pipelines, security checks, and governance controls. GitLab Orbit is being developed to provide a context graph of the software lifecycle, and GitLab Duo Agent Platform orchestrates agent work across stages. The company is also working on attribution and authorization of agent actions and on preserving the instructions and decisions behind code changes.

Availability and what it means

Next-gen SCM is in private beta, and GitLab says it is working with customers to validate performance against their own repositories and workloads. Interested teams can request early access.

GitLab positions the product for the broader enterprise market, in contrast with Scion's narrower focus. Its pitch is agent-scale performance without giving up Git compatibility, infrastructure choice, or integration with the rest of the software development lifecycle. The company also says it will keep contributing to Git and working with the community. For enterprises that cannot move source control onto new infrastructure, the claim is that they can scale for agents without abandoning Git. The performance figures remain GitLab's own early internal results, and independent validation under sustained load is still to come.