GitLab 18.6 release notes

The GitLab MCP server is available in beta. With the GitLab MCP server, you can use AI assistants like Claude Code, Cursor, and other MCP-compatible tools to interact with your GitLab projects, issues, merge requests, and pipelines, all without building custom integrations for each tool.

To get started, turn on beta and experimental features in your GitLab Duo settings.

The GitLab MCP server provides key tools covering issues, merge requests, and pipelines, and we continue to refine it based on user feedback. This feature might have incomplete functionality or bugs. Try it out and share feedback in issue 561564.

Code ownership is critical for maintaining code quality and ensuring the right people review changes to sensitive parts of your codebase. However, managing Code Owners in organizations with complex group structures has been challenging. Previously, to reference a group in your CODEOWNERS file, that group had to be directly invited to each specific project, even if it was already a member of a parent group.

Code Owners now supports groups with inherited memberships as eligible approvers:

• Groups with inherited access through parent group membership are recognized as valid code owners when Code Owners approvals are enabled.
• No need to invite groups directly to every project.
• Existing CODEOWNERS files continue to work without changes.
• Same level of control over who can approve changes to critical code paths.

This change reduces administrative overhead while maintaining the security and approval requirements that Code Owners provide.

On your homepage, draft merge requests can clutter your merge request view and distract from work that’s ready for action. Previously, you could not filter them out.

You can now hide draft merge requests from the Your merge requests section on your homepage by using the display preferences. When you hide draft merge requests:

• They are excluded from the active count.
• A footer displays the number of filtered draft merge requests.
• Your preference is saved automatically.

This change helps you focus on merge requests that need immediate attention.

The GitLab CLI (glab) provides new features and improvements to enhance your GitLab workflow from the command line:

• Enhanced authentication: Auto-detect GitLab URLs from git remotes during login, making it easier to authenticate against the correct GitLab instance.

• Flexible pipeline monitoring: View any pipeline by ID with the ci-view command.

• GPG key management: Manage GPG keys directly from the CLI with new commands.

• Project member management: Add, remove, and update project members from the command line.

• Improved Git integration: Enhanced git-credential plugin with support for all token types.

• Modern user interface: Updated prompt library for better confirmation dialogs and consistent GitLab theme across UI components.

For a full list of changes and updates, see CLI releases. To get started with the GitLab CLI or update to the latest version, see the installation guide.

Webhook integrations are critical for automating workflows and keeping external systems synchronized with GitLab merge request activities. However, when reviewers were re-requested for merge requests, webhook consumers had no way to identify which specific reviewer was being re-requested, making it difficult to trigger appropriate notifications or automation.

Webhook payloads for merge requests now include a re_requested attribute in reviewer data that clearly indicates which reviewer was re-requested:

• Set to true for the specific reviewer being re-requested.
• Set to false for all other reviewers.

This improvement enables more precise automation around the merge request review process. Webhook consumers can send targeted notifications, update external tracking systems, and trigger appropriate workflows when reviews are re-requested.

GitLab Self-Managed administrators in offline or tightly controlled network environments can now configure a custom Web IDE extension host domain, enabling full Web IDE functionality without external internet access.

Previously, the Web IDE required connectivity to .cdn.web-ide.gitlab-static.net to load VS Code extensions and functionality. This requirement blocked Web IDE adoption for security-conscious organizations, government and public sector customers, and enterprises with strict network policies.

With this update, administrators can configure their GitLab instance to serve Web IDE assets directly, removing the dependency on external domains. You can now:

• Use the full Web IDE feature set in completely offline environments.
• Enable the Extension Marketplace with a custom extension registry service.
• Enable markdown preview, code editing, and GitLab Duo Chat within the Web IDE in isolated networks.

Integrating GitLab with external systems through webhooks is critical for automated workflows and keeping teams informed about merge request status changes. However, when GitLab automatically resets approvals (such as when new commits are pushed to a merge request with “Reset approvals on push” enabled), external systems could not distinguish these system-initiated events from manual user actions.

GitLab now includes enhanced webhook payloads that clearly identify system-initiated approval resets. When approvals are automatically reset, webhooks now include:

• A system field set to true.
• A system_action field that provides specific context about why the reset occurred, such as approvals_reset_on_push or code_owner_approvals_reset_on_push.

This means your webhook integrations can now distinguish between manual approval changes and automatic system resets, enabling more sophisticated automation workflows that respond appropriately to the specific context of each approval change.

The GitLab Duo Planner Agent is now available by default in the agent dropdown in GitLab Duo Chat, eliminating the need to manually add it from the AI Catalog. With full context of your work items, epics, issues, and tasks, the Planner Agent can now assist you at both the group and project levels.

Get started with example prompts to see how the Planner Agent can help you break down complex work, create implementation plans, and organize your team’s objectives.

This feature is in beta, and we welcome your feedback in issue 576622.

GitLab’s Helm chart registry previously generated metadata responses on-the-fly, which created performance bottlenecks when repositories contained large numbers of charts. To maintain system stability, we enforced a hard limit of the 1,000 most recent charts. This limit caused frustrating 404 errors when platform teams tried to access older chart versions.

Platform engineers were forced to implement complex workarounds, like splitting charts across multiple repositories, manually managing chart retention policies, or maintaining separate chart storage solutions. These workarounds added operational overhead and fragmented deployment workflows, making it harder to maintain centralized chart governance.

In GitLab 18.6, we’ve eliminated the 1,000 chart limitation by pre-computing metadata responses and storing them in object storage. This architectural change delivers both unlimited chart access and improved performance, as metadata is generated once in background jobs rather than on every request.

Security teams can now use warn mode to test and validate the impact of security policies before applying enforcement, reducing developer friction during security policy rollouts.

When you create or edit a merge request approval policy, you can now choose between warn or enforce enforcement options.

Policies in warn mode generate informative bot comments without blocking merge requests. Optional approvers can be designated as points of contact for policy questions. This approach enables security teams to assess policy impact and build developer trust through transparent, gradual policy adoption.

Clear indicators in merge requests tell users when policies are in warn or enforce mode, and audit events track policy violations and dismissals for compliance reporting. Developers can dismiss vulnerabilities while providing reasoning for the dismissal, creating a collaborative approach to security policy management.

Security teams can now apply business context to projects by leveraging security attributes.

Security attributes are organized by categories including business impact (with structured pre-defined selections), application, business unit, internet exposure, and location. Alternatively, you can create your own attribute categories and define labels within those categories.

By applying these attributes across your projects, you can much more quickly search, filter, and identify which projects within the security inventory that require action based on risk posture and organizational context. You may now:

• Identify projects that are mission critical and requiring better scan coverage
• Review scan coverage by application or business unit
• Search and filter based on the attributes applied to your projects
• Quickly locate projects that contribute to applications which are publicly accessible/exposed

Organizations can now designate specific users, groups, roles, or custom roles that can bypass merge request approval policies in case critical situations occur. This capability provides flexibility for emergency responses, while maintaining comprehensive audit trails and governance controls.

Emergency bypass with accountability: Designated users can bypass approval requirements during critical incidents, security hotfixes, or urgent production issues. When emergencies strike, authorized personnel can merge or push changes immediately while the system captures detailed justification and audit information for compliance review.

Key capabilities include:

• Documented bypass process: When authorized users invoke a policy bypass, they must provide detailed reasoning using an intuitive modal interface, ensuring every exception is properly documented with context.
• Comprehensive audit integration: Every bypass generates detailed audit events including user identity, policy context, reasoning, and timestamps for complete visibility into exception usage patterns.
• Flexible configuration: Define exception permissions for policies using YAML or UI configuration, supporting individual users, GitLab groups, standard roles, and custom roles.
• Git-based push exceptions: Users with pre-approved policy exceptions may push directly when invoking the push bypass option security_policy.bypass_reason.

This feature eliminates the need to entirely disable security policies during emergencies, providing a controlled path for urgent changes while preserving organizational governance and audit requirements.

We’re also releasing GitLab Runner 18.6 today! GitLab Runner is the highly-scalable build agent that runs your CI/CD jobs and sends the results back to a GitLab instance. GitLab Runner works in conjunction with GitLab CI/CD, the open-source continuous integration service included with GitLab.

What’s New:

• Implement minimal job confirmation API

Bug Fixes:

• GitLab Runner does not expand the variables in the Docker image platform option
• Helper sidecar container fails to upload cache to S3 bucket from another account
• Automatically canceled job continues execution and fails
• Missing UTF8 BOM in the generated PowerShell script allows remote code execution using merge request title with character Ä
• Intermittent Kubernetes API server request failures with Kubernetes executor
• When using a Kubernetes executor, jobs with large commit messages fail

The list of all changes is in the GitLab Runner CHANGELOG.