GitLab 19.4 Adds Goal-Driven Agents and Tighter Cost Controls
GitLab 19.4 ships goal-driven agents in Duo CLI—plus hosted open models and MCP tools. Enterprise DevSecOps teams now have native agent capability without external API dependency.
Why it matters
GitLab embeds autonomous agent capability into its core DevSecOps workflow, letting practitioners deploy goal-driven automation inside the platform. The addition of hosted open-weight models and expanded MCP tooling reduces friction for teams balancing cost control with autonomous capability.
The key facts
6 to knowGitLab 19.4 adds goal-driven mode to GitLab Duo CLI
Three hosted open-weight models included (model names not disclosed)
Expanded Model Context Protocol (MCP) tooling set
Enhanced AI spending controls and cost visibility
Agent capability embedded in DevSecOps platform (not bolted on)
Release date: October 12, 2026
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: GitLab has released version 19.4 of its DevSecOps platform, adding a goal-driven mode to GitLab Duo CLI, three hosted open-weight models, a larger set of Model Context Protocol (MCP) tools, and more detailed controls for AI spending. By Matt Saunders