GitLab Introduces AI Governance Dashboard for Agent Visibility
Centralized tracking for internal and external AI coding agents to mitigate shadow AI risks.
1Addressing Shadow AI Risks
The proliferation of AI coding agents like Claude Code, Cursor, and Codex creates significant visibility gaps for organizations. Developers often run these tools locally, leading to shadow AI activity that remains untracked. GitLab's new dashboard addresses this by centralizing activity data, ensuring that every agent interaction is recorded and linked to specific developers.
2Unified Agent Monitoring
The AI Governance Dashboard consolidates data from both the internal GitLab Duo Agent Platform and externally connected agents. By providing a single view, it allows security teams to identify which agents are running, who initiated them, and which projects were modified. This eliminates the need to manually stitch together disparate data sources for audits.
3Quantifying Agent Activity Trends
Three primary tiles at the top of the dashboard provide a baseline of agent performance over seven or thirty-day periods. These metrics track the number of agents, total sessions, and tool calls made. Daily trend analysis helps teams spot anomalies, such as sudden spikes in tool usage that might indicate unauthorized software adoption.
4Detailed Audit Trail Capabilities
Traditional audit logs were designed for human actions, often failing to capture the complexity of AI-driven code changes. GitLab now records every agent action as an audit event, including the specific model used, tools called, and MCP servers involved. This provides an immutable record essential for incident investigation and regulatory compliance reporting.
5Streamlining Compliance and Evidence
When auditors request proof of AI usage, teams can now download session evidence directly from the dashboard. For advanced security needs, organizations can stream these events into their existing security information and event management (SIEM) systems. This capability ensures that accountability is maintained even as development teams rapidly adopt new AI coding tools.
6Configuration and Access Requirements
To utilize the dashboard, users must hold an Owner role or a custom role with read_agent_artifacts permissions at the top-level group. AI audit event storage is disabled by default and must be manually enabled within the Data and privacy settings. This ensures that organizations maintain strict control over their internal data collection processes.
7Integrating External AI Agents
Connecting external agents requires the GitLab CLI version 1.122.0 or later on macOS or Linux. Running the command 'glab govern setup' installs necessary hooks for Claude Code, while additional flags capture Cursor and Codex sessions. A background job transmits these sessions every 30 minutes, ensuring the dashboard remains updated with current developer activity.
8Strategic Importance of Governance
As AI-assisted coding becomes standard, the ability to govern these tools is as critical as the code itself. By providing transparency into agent behavior, GitLab enables organizations to balance developer velocity with necessary security oversight. This beta release marks a shift toward proactive AI risk management within the software development lifecycle.
Original article: GitLab — GitLab Introduces AI Governance Dashboard for Agent Visibility