# GitLab 19.4 introduces per-user AI credit tracking and budget controls Granular visibility and caps for team-based AI spending management - **Source:** [GitLab](https://brifing.si/source/76101668-80db-4240-8676-e968d2b72546) - **Published:** 2026-09-17 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/c73141ed-1791-4ceb-97f1-f76d09508695 ## The AI spending visibility gap While subscription caps control overall AI spend, they lack team-level granularity. Without per-user data, organizations can't allocate fair budgets, explain usage spikes, or demonstrate departmental consumption. GitLab's survey shows 98% of companies have or plan AI governance budgets, highlighting the urgency for better controls. The new system tracks credits by user, namespace, and project with timestamps. ## Granular credit attribution GitLab 19.4 provides detailed tracking where each credit spent is attributed to specific users and teams. The platform records which team consumed credits, on what activity, and when - eliminating the need to reconcile separate systems. Usage data syncs periodically (not real-time), with email alerts at 50%, 80%, and 100% of credit thresholds. This enables both showback reporting and future chargeback capabilities. ## Flexible budget controls Administrators can set default credit caps for all users while maintaining override capabilities. Staff engineers might receive higher limits than contractors, with all adjustments reversible. The system preserves historical override data for future reference. GitLab Flex adds a second control layer, allowing capabilities to be marked as Restricted, Usage cap, or Unlimited against shared reservations. ## Developer visibility features Individual developers can view their daily credit consumption through personal dashboards, with the same filtering capabilities as group views. Promotional credits (12/month for Premium, 24 for Ultimate) are now applied per-user before shared evaluation credits. This ensures users first consume what they're entitled to, while shared pools cover exceptions for heavy users. ## Comprehensive reporting The system generates three export files covering 31-day periods: daily summaries, per-event details, and manifests. Reports include event IDs, timestamps, user/namespace/project IDs, credit quantities, LLM call counts, token totals, and models used. Organizations can map namespaces to financial cost centers for seamless cost structure alignment in monthly reports. ## Implementation workflow Billing managers set subscription ceilings in the Customers Portal. Group Owners configure per-user limits via the Credit caps page, establishing defaults and overrides. Usage data exports generate from either the GitLab Credits dashboard or Flex month overview, delivering event-level detail. Existing automation continues working as the GraphQL API maintains compatibility. ## Platform engineering parallels The approach mirrors CI/CD security evolution - pairing autonomy with visible, enforced boundaries. As agentic automation scales, teams need their own budgets with spending visibility. This enables informed approval of subsequent workflows based on cost transparency, similar to how platform teams made independent CI/CD work safe over the past decade. ## Strategic cost management The system enables organizations to fund heavy AI usage teams without opening budgets to everyone. By setting low default caps and overriding for agentic work users, companies ensure shared pools pay only for exceptions. This granular control becomes critical as AI adoption grows, with GitLab's Transcend event on October 6th exploring how to maintain speed while governing agentic AI across software lifecycles. ## Sources covering this story - https://about.gitlab.com/blog/new-usage-caps-2026/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._