As agents generate more code, engineers’ time and judgment become even more valuable. Route code reviews to the right people the first time with the recommend reviewers flow. Demo below.

I recently sat down with Victor Rodrigues, VP of Customer Software Strategy, Architecture, and Systems at Cochlear, to talk about the software inside Cochlear's hearing implants. Cochlear has used GitLab for 14 years, which keeps its work traceable and auditable. Now AI is changing how the team builds software. GitLab Duo cuts security scanning setup to ~30 minutes instead of two weeks, and Cochlear is testing GitLab Orbit to give AI more context across its large codebase. Thanks, Victor, for the partnership and for sharing Cochlear's story.

Speed and control have to be designed together. GitLab 19.5 brings both to agents, secrets, and code review. Agents in any MCP client can now connect to GitLab and work across code, merge requests, work items, and pipelines, under each user’s permissions. Secrets Manager scopes every credential to the jobs that need it and audits every read. And Recommend Reviewers Flow gets each merge request to the right reviewer sooner.

One of my favorite parts of Transcend was seeing the energy around what comes next for software engineering. We're entering a period where the cost of implementation is falling dramatically, but the opportunity to create value is expanding. At GitLab, our mission is to help organizations turn that opportunity into real outcomes—without giving up control of their IP, their infrastructure, their models, or their security. We announced more than a dozen innovations at Transcend, from agent orchestration and lifecycle context to governed artifacts, dependency security, and AI impact measurement. Some are available today. Others are in beta or coming soon. Together, they represent a clear direction: a governed software factory that can move from intent to production at agent speed. To our customers, partners, and the GitLab team: thank you for helping us build this future. We're just getting started. Explore the announcements and see what you can turn on today: about.gitlab.com/events/tran…

We announced next-gen SCM at Transcend in June, with Nick from Anthropic on stage. Their new post shows why: merges went from 2,500 to 30,000+ a day in four months. They built Scion for their unique needs, enterprises will need that scale on Git, governed end to end, cloud or air-gapped, without migrating their code, CI and compliance workflows to get it. That's what we're building. Latest: about.gitlab.com/blog/scalin…

Enterprises shouldn’t have to surrender control of their most valuable asset, their intellectual property, to benefit from AI. And they shouldn’t have to bet their software engineering future on one model provider. At GitLab, we’re building a cloud-neutral, model-neutral platform for software engineering. That means control over where your code and data live, freedom to choose the best model for each task, and the security and governance needed to trust what AI produces. Sometimes the best choice is a frontier model. Sometimes an open-weight model delivers comparable results at a fraction of the cost. At Transcend, we shared how some open-weight model combinations can deliver up to 8x more calls per GitLab Credit at similar task-completion rates. Our incentives are aligned with our customers: use the right intelligence for the work, at the best price-performance, without sacrificing control. about.gitlab.com/blog/optimi…

You can’t secure a machine-speed software factory with human-speed exception handling. Agents can select dependencies, invoke build tools, access credentials, and publish artifacts far faster than people can manually inspect every action. That means security must become part of the execution path, not a separate process after the work is done. At Transcend, we announced three important building blocks: GitLab Dependency Firewall to enforce policy before risky packages enter the build. GitLab Secrets Manager to scope credentials to the jobs that actually need them. GitLab Artifact Central to govern the software components teams publish and consume. Together, these capabilities help enforce policy where the work happens. The goal isn’t to slow agents down with more gates. It’s to make trusted execution the default. about.gitlab.com/blog/transc…

The most capable AI model in the world is still limited by the context you give it. In enterprise software, that context is scattered across repositories, merge requests, issues, pipelines, security findings, and decisions accumulated over years. Without it, agents guess, repeat work, and make mistakes. That’s why we’re building GitLab Orbit, a context graph for the entire software lifecycle. The early results are compelling: up to 45x fewer retries and 4.5x fewer tokens for agentic workflows. More than 3,500 organizations have already used Orbit during beta. GA coming later this month, try it free while in beta. This is one of the most important architectural investments we’re making. Better context improves accuracy, reduces cost, and helps agents reason about how a change affects the broader system. The future of AI isn’t just smarter models. It’s smarter systems around them. about.gitlab.com/blog/transc…

A coding agent can write the change. But who gets it to production? The code still needs to be reviewed, tested, scanned, approved, and deployed. Those steps often take longer than generating the code itself. And as agents produce more code, the queue only grows. That’s why we’re extending AI beyond coding into the entire software lifecycle. At Transcend, we introduced goal-driven flows in GitLab Duo Agent Platform. Give the system an objective, and agents can coordinate the work across lifecycle stages while respecting your organization’s existing policies and approval gates. This is the difference between automating a task and orchestrating an outcome. We’re building for the outcome. about.gitlab.com/blog/gitlab…

The agent architectures of today will not get us to the autonomous enterprise. For the last decade, humans have orchestrated software delivery across disconnected tools. We move between planning, coding, source control, CI/CD, security, and deployment. Now we’re asking AI agents to do that work. Human tool workflows are becoming machine-executed code paths. That changes everything. Agents aren’t people. They’re nondeterministic systems operating at machine scale. They need explicit identities, bounded privileges, shared context, policy enforcement, and end-to-end observability. If you were designing a distributed system from scratch, would you connect a dozen independent services without these foundations? Of course not. Yet that’s exactly what many enterprises are implicitly asking their agents to navigate today. We call this the shadow software factory. It’s time to replace it with an architecture designed for agents, not just humans. Read more about the building blocks we are putting in place toward a governed software factory that delivers speed you can trust, all the way to production. about.gitlab.com/blog/transc…

We announced Claude Mythos 5 powered scanning and remediation flows with Duo Agent Platform this week, specifically to protect your enterprise by applying frontier models to scanning and remediating your software before this happens to you. Don’t rely on human scale security efforts alone, when hackers are turning to agent-powered exploits at machine speed. DM me for early access, I expect we will begin making this available within the next few weeks.

So excited to debut GitLab Artifact Central. We store your source code, build, test and deploy it. Now we can store and version your artifacts too, reducing pipeline times, build failures and security issues… and save you money in the process. GA coming end of year. Try it now for free!

The autonomous enterprise will require a different architecture than the shadow factory most people are running today. Watch our Transcend event to learn what we’re building and see a dozen new innovations announced including our partnership with Anthropic to bring Claude Mythos 5 vulnerability detection and remediation in Duo Agent Platform.

Another example of what a shadow software factory risks for your business: theregister.com/ai-and-ml/20… Do you really want agents running on your developers’ laptops, under full privileges inherited from their identity, with no policy, no audit log, and no ability to measure ROI? At GitLab Transcend this coming Tuesday, October 6th, I’ll share how you can go fast, and stay in control with a Governed Software Factory running in your cloud and with your choice of models. Register and watch here. about.gitlab.com/events/tran…

I recently sat down with Darren Wilkes, CTO at COCC, to talk about how his team uses @gitlab Duo Agent Platform. Hundreds of AI agents now carry their engineering work from requirement to production, with everything traceable back to source. Watch our conversation

Full lifecycle agentic /goal seeking - now available in Duo Agent Platform CLI. about.gitlab.com/blog/gitlab…