GitLab

Three open weight models are now available and hosted on GitLab Duo Agent Platform, so you can tune quality, latency, and cost per workload, within your guardrails. Learn more in the blog: https://lnkd.in/espM5xgX Open weight models can improve inference cost on many agentic tasks, giving teams up to 4x more calls per GitLab Credit, while keeping performance in line with frontier models. Admins still set the default model per feature and curate what teams can choose from, keeping full control even as the options expand.

A subscription cap keeps total AI spend within budget, but it can't tell you which team spent it to help predict next quarter’s budget needs. Without per-person data, you can't set a fair cap, explain a spike, or show a department what it actually consumed. GitLab Credits usage visibility, now generally available, makes AI spend even more manageable with additional granularity and controls. Set a default cap for every user, override it for individuals, and show developers their own consumption. Export usage down to the billable event, including by user, activity, and timeframe, so you can share usage with teams and departments. Learn more in the blog. https://lnkd.in/efeHV2Rd

Advanced AI models are handing engineering teams something we could only have hoped for a year ago. Building software that tests, inspects, and improves itself with verified remediation at machine speed. GitLab CISO Chaim Mazal breaks down what that takes in practice: know your attack surface, constrain execution, and close the loop from discovery to verified fix. Read his full take as he explains why defenders who build the right operating standard around advanced AI models will turn that advantage into more trustworthy software: https://lnkd.in/eZiTSKBy

The more an agent can do, the more a platform team has to think about what it should be allowed to do, and that's a governance problem that tends to get solved by restricting access rather than by extending it. GitLab 19.4 solves it differently. Read the blog to learn more: https://lnkd.in/ebyWb_Nm New GitLab MCP server tools, now in public beta, let any agent, internal or third-party, run pipelines, open and merge requests, and triage vulnerabilities under the same governance already covering GitLab Duo Agent Platform. Read-only actions default to Always Allow so routine work moves fast, while anything that writes, merges, or deletes waits for the reviewer your team chose. One set of rules, configured once, covers every agent that touches GitLab.

Every AI assistant hits the same wall mid-task, stopping at the end of a turn to wait for you to unstick it. GitLab Duo CLI's new /goal command changes that, taking on a whole objective instead of one turn at a time. Describe the outcome you want, and GitLab Duo CLI makes the changes and runs verification, iterating as needed. A separate model checks that work against your goal at each step, deciding whether it's genuinely done or another iteration is needed. You can stop the run, revise the goal, or step in at any point, so you delegate real work and come back to a completed, verified result instead of a half-finished thread. Read the blog and watch the full demo: https://lnkd.in/eFEjfydZ

As agentic automation spreads from individual developers to entire engineering teams, what limits its reach is no longer what agents can do but how confidently an organization can extend it. GitLab 19.4 brings new agentic automation to every surface developers work in, with new options for cost efficiency and the control platform owners need to scale it. ✨ Highlights include: The /goal command in GitLab Duo CLI, now in public beta, lets developers automate an open-ended objective to a governed agentic flow that runs locally and verifies its own work. Hosted open weight models, now available in GitLab Duo Agent Platform, give teams agentic automation with performance comparable to frontier models, letting teams match model cost to task complexity. New GitLab MCP server tools, now in public beta, let agents automate full tasks across CI/CD, merge requests, work items, vulnerabilities, and projects under the same governance rules as GitLab Duo Agent Platform. GitLab Credits usage visibility, now generally available, gives platform owners the ability to measure and control the cost of agentic automation. Learn more about the 19.4 release: https://lnkd.in/eYMdXFBc Ready to dive deeper into 19.4? The full release blogs are linked in the comments

Small team. Big shipping energy. 👀 For Boundeal, GitLab doesn’t just help its lean team ship like a much bigger one. It’s also the backbone of a broader control environment built to support enterprise customers and certifications like SOC 2 and ISO. See how Boundeal built that foundation from day one in the GitLab Startup Program. 👉 https://lnkd.in/eCeC9YUT

Your AI agent writes code, just never the way your project actually does it. Stop re-explaining your rules every single prompt. With GitLab's Duo Agent Platform, you can give your agent a Skill, one file it reads every time so it already knows your conventions. No classes, no CSS files, no forgotten test files. Just the agent building things your way, automatically. https://lnkd.in/eWYqAjwt Full walkthrough from Ania Kubow.

Agents can write code, review changes, and troubleshoot pipelines, but the handoffs between each step still require manual coordination. Join us at GitLab Transcend to see what we’re building to reduce the bottlenecks between writing code and shipping it. https://lnkd.in/dpz6MWmG Register now to get exclusive access to bonus content from the Enterprise AI Summit livestream, Oct 7-8, a $1,600 value.

Most companies think they have an AI problem when they actually have a fluency problem. Access to tools isn't the same as knowing where and how to apply them, and that gap between teams is only getting wider. In his latest piece, our CIO Manu Narayan breaks down why the old centralized IT model can't keep pace, and makes the case for a hub and spoke approach that pairs platform strategy with AI engineers embedded directly inside business teams. Read the full piece to see how the right operating model turns AI adoption into real advantage. https://lnkd.in/dxugKktz