What changed in GitLab from 18 to 19

Help us learn about your current experience with the documentation. Take the survey.

On August 20, 2026, GitLab 19.3 was released with the following features. We are excited to recognize gkepas as this month’s Notable Contributor! Giannis has made sustained contributions across the GitLab codebase, the contributor platform, client-go, the Terraform provider, and GitLab Orbit. He is also an active translator and a member of the GitLab Community Discord.

Primary features

Flow Creator foundational agent

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue You can now use the Flow Creator, a purpose-built foundational agent in the AI Catalog, to create custom flows for the GitLab Duo Agent Platform through a plain-language conversation. With the Flow Creator, you describe what you want the flow to do, and the agent produces a complete, runnable flow YAML you can register and use immediately. The agent can also help you debug existing flows and explain framework concepts.

Agentic Core

Resolve merge conflicts with GitLab Duo is generally available

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue In previous versions of GitLab, you had to resolve merge conflicts manually in the GitLab UI or from the command line, even for straightforward cases. Now you can ask GitLab Duo to resolve conflicts for you. Start conflict resolution from the merge widget or the Resolve conflicts page. GitLab Duo analyzes the conflicts, edits the files and commits the resolution to the source branch, and then posts a summary comment on the merge request describing what changed.

Resolve review discussions with GitLab Duo is generally available

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue In previous versions of GitLab, to resolve a code review comment, you had to switch to your editor, implement the fix, commit and push the change, and then manually close the thread. Now you can select Resolve with GitLab Duo and GitLab Duo will address the review discussion for you. GitLab Duo reads the comment and the surrounding code, makes the requested change on the source branch, replies to the discussion with a summary of what changed, and then resolves the thread. If the change does not address the comment correctly, you or the reviewer can reopen the thread.

GitLab Duo CLI plugins and marketplaces (Experiment)

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue GitLab Duo CLI now supports plugins and plugin marketplaces as an experiment, introduced in GitLab Duo CLI 9.10.0. A plugin bundles Agent Skills, custom slash commands, and Model Context Protocol (MCP) servers into a single directory. A marketplace is a catalog of available plugins, hosted in a Git repository or a local directory. GitLab Duo CLI automatically registers the official gitlab-duo-plugins marketplace the first time you use plugins. The marketplace includes three skills for common GitLab workflows:
• mr-review: Reviews a merge request and posts comments.
• stack-changes: Splits a large local change into a stacked merge request chain.
• create-issue: Drafts a GitLab issue from a natural-language description. To install one of the skills, run glab duo cli plugin install @gitlab-duo-plugins or duo plugin install @gitlab-duo-plugins, based on your setup. For compatibility with the existing community plugin ecosystem, GitLab Duo CLI also reads .claude-plugin/marketplace.json files, so existing Claude Code plugin marketplaces work with GitLab Duo CLI without modification.

New GitLab Duo Agentic Chat UI in VS Code (Beta)

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue The GitLab Duo Agentic Chat interface in the GitLab for VS Code extension has a redesigned UI, now available as a beta. The new interface is a rebuilt version of the existing Agentic Chat panel, with the same underlying capabilities: AI models, agents, slash commands, and tool approvals. The redesigned UI is available in GitLab 19.3 with GitLab for VS Code extension 6.87.0 and later. To use the new UI, turn on beta and experimental features for your top-level group or instance, then reload the extension. You can switch between the new and classic UI at any time from the panel. The beta is available for VS Code only.

New MCP tools for reading and searching merge requests

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue You can now use get_merge_request to retrieve a merge request along with its diffs, commits, notes, pipelines, or discussions in a single call, so your AI agent no longer has to chain multiple requests to get the full picture of an MR. You can also use the new list_merge_requests tool to search and filter merge requests by author, assignee, reviewer, state, labels, or free-text query, making it easy to find exactly the MRs you care about without leaving your workflow.

Pre-register MCP OAuth applications

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Previously, the mcp scope was hidden from the OAuth applications form in the Admin area, so you couldn’t pre-register an OAuth application for your MCP clients without using Dynamic Client Registration (DCR). Now you can create a shared OAuth application with the mcp scope directly from the Admin area, giving your users a stable client ID to reuse and helping you avoid DCR rate limits on shared networks.

See which user authorized each MCP OAuth application

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue Previously, when MCP clients connected to GitLab using OAuth Dynamic Client Registration (DCR), all dynamically-registered OAuth applications appeared in the Admin Area with only a generic client name, making it impossible to tell which user authorized a given application. Now, when you approve an MCP OAuth connection, your username is automatically appended to the application name — for example, [Unverified Dynamic Application] kiro — authorized by @username. You can quickly identify which user is behind each dynamic OAuth application directly from the Admin Area, without any additional configuration.

View agent task plan in session detail sidebar

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Previously, when you opened a session detail page, you had to manually expand individual activity log entries to find the agent’s current task list. Now, the session detail sidebar shows the agent’s latest plan automatically. You can view each task and its status, and a progress summary like “2 of 5 completed.” You can track what the agent is working on and how far along it is without digging through the activity log.

Devstral 2 deprecated for GitLab Duo Agent Platform Self-Hosted

• Tier: Premium, Ultimate
• Offering: GitLab Self-Managed
• Links: Documentation · Related Issue Mistral has deprecated the Devstral 2 model, and GitLab has deprecated Devstral 2 for GitLab Duo Agent Platform Self-Hosted. Mistral Medium 3.5 is a supported alternative available from Mistral.

Support for Amazon Bedrock Mantle (Beta)

• Tier: Premium, Ultimate
• Offering: GitLab Self-Managed
• Links: Documentation · Related Issue If you use GitLab Self-Managed with self-hosted models, you can now use Amazon Bedrock Mantle as a supported provider. Mantle is an OpenAI-compatible inference engine for Amazon Bedrock, so you can use familiar OpenAI SDKs and integrations with Bedrock models. This feature is in beta.

Unified DevOps and Security

Enforce merge trains

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue In previous versions of GitLab, you couldn’t stop merges from bypassing the train. Both the option to merge immediately and the REST API could skip merge train protections without restriction. For teams running high-velocity monorepos, a single merge that skips the train can cancel and restart every in-progress pipeline, multiplying CI costs and straining infrastructure. Now you can enforce merge train usage across the UI and API with a single project-level setting, preventing bypasses that cancel and restart in-progress pipelines. Owners and Administrators can still override the setting when needed.

GitLab Secret Scanning for Source Code (Beta)

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed
• Links: Documentation · Related Issue GitLab Secret Scanning for Source Code is now in beta, powered by a new GitLab-built scan engine. Unlike the default analyzer, which detects only known secret patterns, this analyzer also detects passwords and other unstructured secrets that fall outside standard ruleset coverage. It also uses multiple heuristic techniques to reduce false positives. The new analyzer replaces the default analyzer in the same secret_detection job, matching existing vulnerability findings instead of creating duplicates. To get started, see turn on the analyzer. During beta, only high-confidence findings are reported. We welcome any feedback you have in issue 609578.

Secret detection scans commit history on default branch pushes

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue Secret detection on the default branch now scans all commit diffs in a push when a prior commit reference is available, rather than scanning only the latest directory contents. This change closes a gap where secrets introduced and removed within the same push went undetected. The behavior now aligns with how secret detection works on merge requests and feature branches. This scanning catches secrets that briefly existed in your repository history even if removed before the pipeline completes. Security teams can now identify secrets that were ever committed, not just those present at HEAD. For more information, see pipeline secret detection coverage.

Disable OAuth Dynamic Client Registration for MCP

• Tier: Free, Premium, Ultimate
• Offering: GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Previously, MCP clients and AI tools could automatically register OAuth applications on your instance through Dynamic Client Registration (DCR), which you couldn’t turn off. This made it difficult for administrators on GitLab Self-Managed and GitLab Dedicated instances to control which OAuth clients could connect. Now you can disable DCR entirely using the application settings API, giving you full control over which OAuth clients can access your instance. When DCR is disabled, clients must use a pre-registered OAuth application instead of registering automatically.

SPDX license expressions in CycloneDX SBOMs brought to GitLab

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue GitLab 19.3 adds support for Software Package Data Exchange (SPDX) license expressions in CycloneDX Software Bill of Materials (SBOM) files you bring to GitLab. Previously, components with composite or custom licenses defined using SPDX expression syntax would appear as unknown. …

Help us learn about your current experience with the documentation. Take the survey.

On July 16, 2026, GitLab 19.2 was released with the following features. We are excited to recognize vivekshukl007 as this month’s Notable Contributor! Vivek is a Senior Data Engineer at Emirates who made a high-impact contribution to add Scala language support to GitLab Orbit. This change improves the code graph capabilities for developers who work in Scala codebases.

Primary features

GitLab Duo CLI is now generally available

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue GitLab Duo CLI brings the GitLab Duo Agent Platform directly to your terminal. Use the CLI to ask complex questions about your codebase and to autonomously perform actions on your behalf. Unlike external tools, the CLI has context about your GitLab project, pipelines, and agent configurations. Key features include:
• Two modes: interactive chat mode and headless mode for CI/CD
• Administrator on/off control for GitLab Self-Managed and GitLab Dedicated
• Model selection and shared sessions
• Tool approvals
• Model Context Protocol (MCP) connections
• Slash commands, including commands for context usage and context compaction
• Support for skills and AGENTS.md customization files Install the GitLab Duo CLI through the GitLab CLI (glab) or as a standalone tool.

GitLab Duo custom flows are now generally available

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Custom flows are AI-powered workflows you create and configure to automate complex, multi-step tasks across your GitLab projects. They let teams define workflow steps, components, and triggers so repetitive development and operational work can run automatically in response to GitLab events. In the GitLab UI, flows run directly in GitLab CI/CD, helping teams automate common tasks without leaving GitLab. Key features include:
• YAML-defined, reusable workflows for team-specific automation
• Multi-agent orchestration for complex, multi-step tasks
• User-defined human-in-the-loop (HITL) checkpoints for approval or feedback at sensitive steps
• Native GitLab triggers, including mentions, assignments, pipeline events, and merge request lifecycle events
• Flow creation and management from projects or the AI Catalog
• Public and private visibility controls
• Secure execution using service accounts and composite identity
• YAML validation to catch configuration issues before runtime

Scheduled pipeline execution policies are GA

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Scheduled pipeline execution policies are now generally available. Define a schedule once in a security policy project and enforce it across every project in scope, without editing each project’s .gitlab-ci.yml. If requirements change, update the policy in one place instead of coordinating changes across many CI/CD configuration files. Use scheduled policies to run compliance scripts, security scans, or other custom CI/CD jobs on a daily, weekly, or monthly cadence, independent of commit activity. This is useful for repositories without regular code changes, such as running dependency scans to detect newly discovered vulnerabilities. Each policy runs as a separate pipeline, with time zone support, time window distribution, and branch targeting.

Start foundational flows from Agentic Chat

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue In previous versions of GitLab, you started foundational flows from specific UI actions, mentions, or assignments. Now you can start them from Agentic Chat in the GitLab UI as part of your conversation. When your request matches a specialist workflow, Agentic Chat hands off to one of these flows:
• Developer Flow: Implements changes or opens a merge request
• Code Review Flow: Reviews a merge request
• Fix CI/CD Pipeline Flow: Diagnoses and repairs a failed pipeline You approve the handoff in chat, then follow progress in the conversation or from AI > Sessions.

Dependency scanning auto-remediation (Beta)

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue GitLab 19.2 introduces Dependency scanning auto-remediation in Beta. The feature brings automated vulnerability remediation directly into your dependency scanning workflow, with two capabilities:
• Automated dependency version bumps, available on GitLab.com, GitLab Self-Managed, and GitLab Dedicated.
• Agentic Breaking Change Resolution, available on GitLab.com, GitLab Self-Managed, and GitLab Dedicated, and consumes GitLab Credits. Automated dependency version bumps automatically opens merge requests to update vulnerable dependencies to their safe versions. Once turned on, GitLab monitors your projects for vulnerable dependencies and opens remediation MRs without manual intervention. By default, updates target patch and minor versions. Agentic Breaking Change Resolution extends the remediation flow to handle complex updates. When a merge request that bumps dependency versions has a pipeline fails on a breaking change, GitLab Duo analyzes the pipeline errors, the dependency’s changelog, and how your code uses the dependency. GitLab Duo commits fixes to the same MR and re-runs the pipeline until the pipeline passes. When you enable Agentic Breaking Change Resolution, version bumps extend to include major versions. Together, the two capabilities form a complete remediation loop: GitLab opens the MR, and when the update is complex, GitLab Duo resolves it. For setup instructions, see Dependency scanning auto-remediation. Share feedback in the beta feedback issue.

Non default branch tracking (beta)

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue You can now track vulnerabilities on branches other than the default branch. For the best results, target a small number of long-lived release branches, such as branches for specific environments (project-qa, project-prod) or deployment platforms (project-iOS, project-android). This beta includes the following capabilities:
• Add tracked branches on the security configuration page, up to twice the number of projects in the namespace.
• Filter by branch on the vulnerability report.
• Filter by branch on the project-level security dashboard.
• Track all vulnerability types on tracked branches, including CVEs, which were previously out of scope.
• Keep vulnerability status metadata consistent when a branch merges into the default branch.
• Update vulnerability status on tracked branches.

Selective GitLab Duo availability for subgroups

• Tier: Ultimate
• Offering: GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue Administrators of GitLab Dedicated instances can make GitLab Duo and GitLab Duo Agent Platform unavailable for selected subgroups while other subgroups still have the option to turn them on. Previously, you could either disable GitLab Duo and Agent Platform for an entire instance, or make them potentially available for all. Now you can enforce a default-deny, per-subgroup allowlist. Mark specific subgroups as Always off (locked) so their descendant groups and projects can never enable GitLab Duo and Agent Platform, while leaving other subgroups up to the discretion of users with the Owner role. Only administrators can apply or remove the lock, and affected Owners see clear messaging that GitLab Duo is locked by a parent group. This feature helps compliance and platform governance teams meet strict data-classification requirements.

AI audit event report (beta)

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue AI audit event reports are now available in beta, giving security and compliance teams a unified, downloadable record of GitLab Duo agent activity. Previously, agent activity was scattered across pipeline jobs and event histories, making it difficult to reconstruct a session for:
• Incident investigation.
• Compliance review.
• AI governance reporting. Now, each agent session produces a comprehensive audit artifact capturing:
• Inputs.
• Model and configuration context.
• The chronological event timeline.
• Outputs. You can browse AI audit events from the Governance page, filter by agent and session details, drill into individual events, and download the underlying session artifact.

Security Review Flow (beta)

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Security Review Flow detects business logic vulnerabilities directly in merge requests. Unlike static analysis tools that scan for known patterns, Security Review Flow reasons about the intent of your code and identifies authorization bypasses, data exposure, and logic errors that pattern-based scanners routinely miss. To request a review, assign the Duo Security Review service account as a reviewer on your merge request. The flow analyzes the diff and posts findings as threaded comments at the exact lines where vulnerabilities occur, each with a Common Weakness Enumeration (CWE) classification, severity rating, and where possible, an inline suggested fix you can apply without leaving the merge request. Each review consumes GitLab Credits based on the complexity of the merge request diff.

Agentic Core

Bulk AI Catalog items enablement

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue When enabling public custom and external agents and public custom flows in the AI Catalog, you can now select up to 100 projects in a single action, instead of selecting individual projects one at a time.

Configure ID tokens in flows

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue Use ID tokens to authenticate with third-party OpenID Connect (OIDC) services without storing long-lived credentials. For example, use ID tokens for keyless signing of binaries and commits, or to retrieve secrets from a secrets manager. To use this feature, update your agent configuration to include the id_tokens keyword, then configure the service to trust tokens issued by GitLab Duo Agent Platform.

Custom Agent validation

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Previously, you could save a custom agent in AI Catalog whose prompt would fail when run. For example, prompts that tripped security rules caused the agent to silently do nothing when being used. Now, when you create or update a custom agent, GitLab validates the prompt configuration, and tells you about any errors before you save that agent.

Turn on MCP server independently from the Agent Platform

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue To give you finer control over how external tools connect to your GitLab instance or group, you can now turn the GitLab MCP server on or off independently from Agent Platform settings. …

Help us learn about your current experience with the documentation. Take the survey.

On June 18, 2026, GitLab 19.1 was released with the following features. We’d also like to announce this month’s Notable Contributor: Pishel65! We are excited to recognize Pishel65, a Level 3 contributor with 19 merged MRs and 9 more open since joining in October 2025.

Primary features

Secret false positive detection with GitLab Duo

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue Secret false positive detection with the GitLab Duo Agent Platform is now generally available. Security teams spend significant time investigating secret detection findings that are incorrectly flagged as actual secrets. These false positives create alert fatigue, erode trust in scan results, and divert attention from genuine security risks. When a security scan runs, GitLab Duo automatically analyzes each critical and high severity secret detection vulnerability to determine if it is a false positive. The AI assessment appears in the vulnerability report, so you have immediate context for faster and more confident triage decisions. Key features include:
• Automatic analysis: Runs after each security scan without manual trigger.
• Manual trigger: Trigger false positive detection for individual vulnerabilities on the vulnerability details page for on-demand analysis.
• Focus on high-impact findings: Analyze only critical and high severity vulnerabilities to maximize signal-to-noise improvement.
• Contextual AI reasoning: Each assessment includes an explanation of why the finding is likely a true positive, based on code context and vulnerability characteristics.
• Confidence scoring: Each detection includes a confidence score to help teams prioritize review based on the model’s certainty.
• Seamless workflow integration: Results appear directly in the vulnerability report alongside existing severity, status, and remediation information. We welcome your feedback in issue 592861.

Always on availability mode for GitLab Duo

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue Administrators can now set GitLab Duo to be always on for all projects in an entire instance or top-level group. When GitLab Duo is set to always on, group, subgroup, and project owners cannot turn off GitLab Duo, giving enterprises centralized AI governance for compliance and regulated environments. This new setting is symmetrical to the existing always off setting, closing a gap where GitLab Duo could be locked off but could not be locked on. This new setting is especially valuable for organizations with autonomous divisions or subsidiaries that need to guarantee consistent AI tooling across the business. To set GitLab Duo to be always on, go the instance or top-level group GitLab Duo settings and set GitLab Duo availability to Always on.

Automatically assign Code Owners as reviewers

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Previously, you needed to select reviewers for each merge request manually, even when a CODEOWNERS file already defined who should review each file. You can now configure a project to assign Code Owners as reviewers automatically. GitLab assigns every Code Owner that matches the changed files. This happens when a merge request is created in a ready state, or when a draft is marked ready. If you already assigned a reviewer, GitLab skips automatic assignment and keeps your choice. To turn on automatic reviewer assignment, go to Settings > Merge requests > Automatic reviewer assignment and select Automatically assign all code owners as reviewers.

Compliance framework templates (beta)

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue You can now create compliance frameworks from predefined templates. Previously, building a compliance framework required defining every requirement and control by hand, a repetitive process when a framework had dozens of controls. Now, when you create a new framework in the Compliance center, you can:
• Select Create from template to start from a preconfigured framework with requirements and controls already in place.
• Preview each template, customize the name, description, and color, and then apply it to your group in a single step. 19 templates are available, including ISO 27001:2022, SOC 2, FedRAMP, NIST, CIS, TISAX, and more.

Improved secret detection coverage for feature branch pipelines

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue In GitLab versions earlier than 19.1, you couldn’t trust a feature branch pipeline to surface every secret in your branch. A new branch scanned only the latest commit. An existing branch scanned only your most recent push. A credential leaked in an earlier commit could sit undetected, reaching shared branches or production before being flagged. Now you can catch those secrets where they’re cheapest to fix. In GitLab 19.1, secret detection scans every commit from the branch’s divergence point with the default branch to the latest commit. That means fewer secrets slip through to later stages, less time rotating exposed credentials after the fact, and consistent, predictable coverage across your branches.

Tool approval guardrails for GitLab Duo agents (beta)

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated, GitLab Dedicated for Government
• Links: Documentation · Related Issue Administrators can now configure tool-level approval policies for GitLab Duo agents, gating sensitive actions with human approval at the moment of execution. Previously, after an AI agent was approved for a project, it could invoke any of its tools without further review, including write and destructive operations. Now, you can define rules for groups and projects that map each tool to one of three modes:
• Allow (execute silently).
• Ask (require human approval).
• Deny (block entirely). When an AI agent calls a tool in “ask” mode, the user is prompted with an inline approval card before execution proceeds. This beta release includes Agentic Chat, IDE, and flows, and emits audit events for every approval decision.

Agentic Core

Custom and external AI feature controls

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Administrators and top-level group Owners can now control which AI agents and flows are available within their organization. They can:
• Prevent users from creating or enabling custom agents and flows. This ensures only centrally-approved AI automation is in use.
• Restrict users from enabling agents and flows owned by projects outside the group hierarchy. This limits exposure to unapproved external content.

Custom flows YAML validation

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue The AI Catalog now validates your custom flow configuration before saving or triggering it. Previously, syntax errors and misconfigured parameters in a custom flow (for example, missing inputs or unknown tool parameters) only surfaced at runtime, after a CI job had already started. This made debugging slow and difficult. Now, when you save or update a custom flow in the AI Catalog, GitLab checks the configuration upfront and surfaces any errors directly in the UI. Valid flows are unaffected and continue to save and trigger as usual.

Pattern-based tool approval for Agentic Chat

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue This feature was removed on July 10, 2026. Previously, when Agentic Chat asked you to approve a tool invocation, you could approve it once or approve the tool call with these arguments for the remainder of the session. Different arguments would require additional approval. Workflows that repeated similar commands, such as a series of git operations, forced you through a stream of nearly identical prompts. You can now choose a third approval option, Approve all uses of this tool for session. This option approves invocations of the tool for the remainder of the session whenever the arguments match the approved pattern. Pattern-based approvals are available for Agentic Chat in the GitLab UI, GitLab Duo CLI, GitLab for VS Code, and the GitLab Duo plugin for JetBrains IDEs.

Automatic code review for new GitLab customers

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com
• Links: Documentation · Related Issue Automatic reviews in Code Review Flow are now turned on by default for new GitLab Duo trial customers on GitLab.com, so you can start getting AI-powered feedback on your merge requests from day one — without any manual setup. With a new flat pricing model, you get immediate value from smarter, faster code reviews right out of the box. If needed, you can opt out in your group settings.

Foundational flows readiness checks

• Tier: Premium, Ultimate
• Offering: GitLab Self-Managed, GitLab Dedicated for Government
• Links: Documentation · Related Issue GitLab Duo health checks now include foundational flows readiness checks, which verify:
• The instance-level flow execution setting is enabled.
• The instance-level foundational flows setting is enabled.
• At least one active instance runner with the gitlab--duo tag is registered and connected, and uses a Docker-compatible executor.

GPT models for Code Review Flow

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue In previous versions of GitLab, Code Review Flow supported only Anthropic Claude models. Teams that could not use Anthropic models due to contractual, policy, or procurement constraints had no way to run Code Review Flow. You can now select GPT-5.2 or GPT-5.3 Codex as the model for Code Review Flow. Top-level group Owners can switch the model for Agentic Code Review in Settings > GitLab Duo > Configure features, under GitLab Duo Agent Platform. The GPT models are hosted through the GitLab AI Gateway, so no additional configuration is required. Both models passed benchmark evaluation against the GitLab Duo code review dataset, with review quality comparable to the default Claude Sonnet 4.6 Vertex model. See the code review benchmark for results.

Model selection allowlist

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue For GitLab Duo Agentic Chat, you can now configure an allowlist of approved models, and set an organization-wide default, if you are:
• On GitLab.com, a top-level group Owner.
• On GitLab Self-Managed, an instance administrator. This gives organizations control over which models users can select when using Agentic Chat.

New event triggers for flows and external agents

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue In previous versions of GitLab, you could only run flows and external agents when the service account was mentioned, assigned, or added as a reviewer. Coordinating automation around the rest of the merge request lifecycle, or around work item creation, required external glue. You can now configure triggers for four additional events:
• Merge request ready: A user marks a draft merge request as ready for review. Previously released behind a feature flag, this event trigger is now generally available. …

Help us learn about your current experience with the documentation. Take the survey.

On May 21, 2026, GitLab 19.0 was released with the following features. We’d also like to announce this month’s Notable Contributor: Norman Debald! We are excited to recognize Norman, a Level 3 contributor with more than 40 merged improvements across GitLab since joining in May 2022.

Primary features

Group-level custom review instructions for GitLab Duo

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Add-ons: GitLab Duo Enterprise
• Links: Documentation · Related Issue In previous versions of GitLab, you could only define custom review instructions for GitLab Duo at the project level. Teams working across many projects in the same group had to duplicate the same instructions in every project. Now you can configure shared custom review instructions for an entire group and its subgroups. Select a project in your group to use as a template. When GitLab Duo performs a code review, it combines the group-level .gitlab/duo/mr-review-instructions.yaml file with any instructions defined in the individual project. Both Code Review Flow and GitLab Duo Code Review support group-level custom instructions.

Configure work item types

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Previously, work item types could be either an Issue or a Task. You can now configure custom work item types in a project to match the way your team plans and tracks work. You can create or rename types to User Story, Bug, or Maintenance. Each work item displays with its type name and a unique icon. The new types support custom fields and status lifecycles, and appear in your saved views and issue boards. Type configuration in the top-level group (GitLab.com) or organization (GitLab Self-Managed) cascades down to all projects. You can also control which types are available for each project. Enable or disable a type across all projects at once, or let individual projects manage their own type visibility. When you disable a type in a project, existing work items are not affected.

GitLab Secrets Manager now available in open beta

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed
• Links: Documentation · Related Issue In previous versions of GitLab, the GitLab Secrets Manager was available only to a closed beta cohort. Most teams relied on external services such as HashiCorp Vault or AWS Secrets Manager. The GitLab Secrets Manager is now available in open beta for Premium and Ultimate customers on GitLab.com and GitLab Self-Managed. When the GitLab Secrets Manager is enabled, project and group Owners can store, retrieve, and reference CI/CD secrets in GitLab. Secrets are scoped to a project or group and are accessible to only pipeline jobs that explicitly request them. During open beta, GitLab Secrets Manager follows the beta support policy and might not be ready for production use. To share feedback, see issue 598100.

GitLab Duo Developer enhancements for merge request workflows

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue GitLab Duo Developer now supports multiple trigger methods: assign it to an issue, select Generate MR, or @mention it in any issue or MR discussion thread to turn feedback, To-do items, and design questions into code changes, follow-up MRs, or research summaries. With AGENTS.md and agent-config.yml configured, GitLab Duo Developer runs your tests and checks before committing. After a top-level group or instance administrator enables the Developer Flow, GitLab automatically adds mention and assign triggers to eligible projects.

Dependency scanning by using SBOM generally available

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue The GitLab SBOM-based dependency scanner is now generally available. Maven, Gradle, and Python projects now have complete visibility into vulnerabilities across their full dependency tree, including vulnerable packages introduced transitively, not just those declared directly. The analyzer now includes automatic dependency resolution for Maven, Gradle, and Python projects. When a lockfile or resolved dependency graph is not present, the analyzer automatically invokes tooling to resolve the full transitive dependency graph before scanning. Dependency resolution is enabled by default and requires little-to-no additional configuration beyond including the v2 Dependency Scanning template. For projects where dependency resolution is not possible, the analyzer falls back to manifest scanning. It parses pom.xml, requirements.txt, build.gradle, and build.gradle.kts to identify direct dependencies. Manifest scanning ensures teams always get a starting point for vulnerability coverage, even for projects without lock or build files. Manifest scanning is enabled by default and returns direct dependencies only. For full transitive coverage, enable dependency resolution or provide a dependency lockfile or graph export manually.

Agentic Core

GitLab Duo Core moves to usage-based billing

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Starting in GitLab 19.0, GitLab Duo Core moves to usage-based billing. Code Suggestions in the Web IDE and desktop IDEs now consume GitLab Credits. GitLab Duo Chat is also changing. For GitLab Duo Core users, Chat is now agentic and runs on GitLab Duo Agent Platform. To use GitLab Duo Chat in the GitLab UI or desktop IDEs, enable GitLab Duo Agent Platform for your instance or top-level group.

Filter exact code search results by repository

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed
• Links: Documentation · Related Issue You can now filter exact code search results by repository. With the repo: syntax, you can directly scope your search query to specific repositories or repository patterns without having to go to individual projects. For example, searching for def authenticate repo:my-group/my-project returns results only from that repository. You can also use partial paths or patterns to match multiple repositories.

Merge request ready event trigger

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed
• Links: Documentation · Related Issue You can now configure flows and external agents to run on the Merge request ready event. When a draft merge request is marked as ready for review, GitLab Duo automatically runs the flow or external agent. To configure a trigger, go to AI > Triggers in your project. This feature is behind the merge_request_ready_flow_trigger feature flag, disabled by default.

Claude Opus 4.7 now available in GitLab Duo Agent Platform

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Claude Opus 4.7 is now available in GitLab Duo Agent Platform. Opus 4.7 delivers meaningful improvements to complex, multistep tasks that require sustained reasoning, precise instruction following, and self-verification before surfacing results. This includes flows supporting CI/CD pipelines, code review, vulnerability resolution, and more.

Support for self-hosted Gemini models

• Tier: Premium, Ultimate
• Offering: GitLab Self-Managed
• Links: Documentation · Related Issue GitLab Duo Agent Platform Self-Hosted is now compatible with Gemini models. Gemini models support multiple flows, including the Code Review Flow, SAST Vulnerability Resolution Flow, Fix CI/CD Pipeline Flow, and more.

Expanded open source model support in GitLab Duo Agent Platform

• Tier: Premium, Ultimate
• Offering: GitLab Self-Managed
• Links: Documentation · Related Issue GitLab Duo Agent Platform now supports additional open source models for self-hosted deployments, including Devstral 2 123B, GLM-5.1-FP8, and others. This helps customers power agentic workflows across a variety of environments, including offline and network-restricted deployments.

Per-session tool approvals with admin controls

• Tier: Free, Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Before GitLab Duo Agentic Chat can use a tool on your behalf, it requires your approval. Each tool invocation requires a separate approval. Now, you can approve a trusted tool once for an entire session and streamline your workflows. Administrators control whether tool approval for sessions is available. The following settings cascade from instance to group to project:
• On by default
• Off by default
• Always off Groups and subgroups can modify the setting unless an administrator sets it to Always off. The default setting is Off by default, ensuring each tool invocation requires explicit approval unless an administrator changes it.

Resolve merge conflicts with GitLab Duo (Beta)

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue In previous versions of GitLab, you had to resolve merge conflicts manually in the GitLab UI or from the command line, even for straightforward cases. Now GitLab Duo can autonomously analyze merge conflicts, edit the conflicting files, create a commit, and push to the source branch. Trigger conflict resolution from the Resolve conflicts page or directly from the merge request widget. When complete, GitLab Duo posts a summary comment so reviewers can see what changed. GitLab Duo respects branch protection rules and does not force-push to protected branches. This feature is in beta and is gated behind the mr_ai_resolve_conflicts feature flag, enabled by default.

Restrict the AI Catalog to a group hierarchy

• Tier: Premium, Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Top-level group Owners can now restrict the AI Catalog to show only agents and flows owned by projects within their group hierarchy. This blocks agents, external agents, or flows not in this hierarchy from being visible or enabled by any user in that group.

Purchase credits on the Free tier on GitLab Self-Managed

• Tier: Free
• Offering: GitLab Self-Managed
• Links: Documentation · Related Issue Free tier users on GitLab Self-Managed can now unlock the full power of GitLab Duo Agent Platform, no Premium or Ultimate subscription required. Choose your monthly credit amount, commit to an annual term, and get instant access to AI-powered development tools. Credits refresh automatically each month, so your team always has what it needs to build faster and smarter.

Admin-defined network access controls for Agent Platform remote flows

• Tier: Ultimate
• Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
• Links: Documentation · Related Issue Administrators can now define centralized network policies for GitLab Duo Agent Platform remote flows directly in Settings. Top-level group administrators on GitLab.com, and instance administrators on GitLab Self-Managed and Dedicated, can configure organization-wide domain denylists and allowlists that projects inherit automatically. An additional setting controls whether projects can extend the approved domain list with custom entries. Policies are enforced at runtime across all remote flows, giving security and platform teams a consistent governance layer for agent network egress.

Unified DevOps and Security

Auto remediation for vulnerable dependencies (Experiment)

• Tier: Ultimate
• Offering: GitLab.com
• Links: Documentation · Related Issue …