GitLab 19.4: Agentic Automation at Lower Cost

Engineering leaders want to scale AI agents across DevSecOps without losing control of cost, permissions or audit. GitLab 19.4 answers that challenge by embedding agentic automation into the same platform that already runs the code, so every action inherits existing rules and every credit is traceable to a user.

What was announced

GitLab Inc. released GitLab 19.4, adding agentic automation across the developer terminal, merge requests and third-party MCP clients. The release introduces the /goal command in GitLab Duo CLI (public beta), new GitLab-hosted open weight models in GitLab Duo Agent Platform, expanded GitLab MCP server tools (public beta), and general availability of GitLab Credits usage visibility.

According to GitLab, the new hosted open weight models Kimi K3, MiniMax M3 and GLM 5.3 deliver up to 4x more calls per GitLab Credit than many comparable frontier models, giving teams a direct lever to balance quality, latency and cost per workload.

Why this matters

For CIOs, CISOs, IT directors and procurement leaders, the limit on agentic AI is rarely capability. It is governance, unit cost and accountability. GitLab 19.4 addresses all three: the same permissions that cover the code govern the agents, tool-level rules apply to third-party MCP clients, and per-user credit attribution lets platform owners widen access without an open-ended risk on the invoice.

This release takes agentic automation from something individual developers use to something an organization can scale at speed and under the controls already in place

Technical details

• /goal command: in GitLab Duo CLI (public beta) automates an open-ended objective locally, with a separate model verifying each step against the stated goal.
• Hosted open weight models: Kimi K3, MiniMax M3 and GLM 5.3 in GitLab Duo Agent Platform, up to 4x more calls per GitLab Credit versus comparable frontier models.
• MCP server tools: public beta tools to trigger pipelines, read failed job traces, run merge requests end to end, search and update work items and triage vulnerabilities.
• Governance defaults: read-only tools default to Always Allow, write and delete tools default to Always Ask.
• Credits visibility: per-user caps, billable-event usage exports by email, developer-level consumption view for GitLab Flex and non-Flex subscriptions.
• Additional features: Duo Agent Platform in Slack (experiment for Premium and Ultimate), model selection for the Developer Flow (GA), redesigned session details panel and Co-Create community MCP tools.

Softprom and GitLab

Softprom is the official partner of GitLab. Our team helps enterprises evaluate, license and roll out GitLab Duo Agent Platform with the right governance, credit controls and integration into existing DevSecOps pipelines.

This content was prepared as part of the Softprom DistriFlow project — an automated system for monitoring and adapting vendor news. Original source: original article.