Skip to main content

overview

GitLab Duo - AI-Powered DevSecOps Platform

Overview

GitLab Duo is GitLab's comprehensive AI-powered assistant that accelerates software development across the entire DevSecOps lifecycle. Starting with GitLab 18.0 (released May 2025), Duo Code Suggestions and Duo Chat are included with Premium and Ultimate tiers at no additional cost.

What is GitLab Duo?

GitLab Duo is an AI-powered platform that provides:

  • Code Assistance: Intelligent code completion, generation, and suggestions
  • Chat Interface: Conversational AI for explanations, refactoring, and test generation
  • Security Analysis: Automated vulnerability detection and remediation
  • CI/CD Intelligence: Pipeline failure analysis and automated fixes
  • Planning Support: Issue description generation and work breakdown
  • Code Review: Automated reviews and suggested reviewers
  • Agent Platform: Orchestration layer for multi-agent workflows

Key Components

1. GitLab Duo Code Suggestions

Real-time, contextual code completions and full code block generation powered by advanced LLMs.

Current Model (as of GitLab 18.2):

  • Primary: Claude Sonnet 4 (default)
  • Previous: Codestral (via Fireworks, GitLab 18.1)

Features:

  • Line-by-line code completion
  • Multi-line code block generation
  • Context-aware suggestions
  • 14+ programming languages supported
  • IDE integration (VS Code, JetBrains, Visual Studio, NeoVIM, GitLab Web IDE)

2. GitLab Duo Chat

Conversational AI assistant for development tasks.

Capabilities:

  • Explain code and errors
  • Generate tests
  • Refactor code
  • Fix issues
  • Answer GitLab feature questions
  • Research project history

Availability:

  • GitLab UI
  • Web IDE
  • VS Code
  • JetBrains IDEs
  • Visual Studio

3. GitLab Duo Agent Platform

General Availability: January 2026 (GitLab 18.8)

Multi-agent orchestration platform for complex DevSecOps workflows.

Key Features:

  • Multi-agent parallel execution
  • Custom workflows (Flows) defined in YAML
  • Pre-built flows for common tasks
  • Integration with external systems via Model Context Protocol (MCP)
  • Full project context awareness

Built-in Flows:

  • Software Development Flow (Beta): End-to-end feature implementation
  • Issue-to-MR Flow: Automatic merge request creation from issues
  • Convert CI File Flow: Migrate CI/CD configurations to GitLab format
  • Planner Agent: Issue planning with create/edit capabilities (Beta)

4. GitLab Duo AI Catalog

Central library for sharing and collaborating on custom-built agents across organizations.

Tier Availability

GitLab Premium

Included (as of GitLab 18.0):

  • Code Suggestions
  • Duo Chat
  • Merge Request Summaries
  • Code Explanation

GitLab Ultimate

Additional Features:

  • GitLab Duo Agent Platform
  • Custom Agents
  • Root Cause Analysis
  • SAST False Positive Detection
  • Vulnerability Explanations
  • Suggested Reviewers
  • Advanced Security Features

GitLab Duo Pro

Standalone Add-on: $19/user/month

  • Available for organizations without Premium/Ultimate
  • Full Duo features including Code Suggestions and Chat

Supported Languages

Code Suggestions supports 14+ programming languages:

  • C/C++
  • C#
  • Go
  • Java
  • JavaScript/TypeScript
  • Kotlin
  • PHP
  • Python
  • Ruby
  • Rust
  • Scala
  • Swift
  • And more

Model Selection

Available: GitLab 18.4+ (GA in 18.5)

Organizations can select AI models for specific features at the top-level group:

Supported Model Providers:

  • Anthropic (Claude)
  • Mistral AI
  • OpenAI (GPT)
  • Codestral

Model Selection Features:

  • Group-level configuration
  • Cascades to child groups and projects
  • Feature-specific model assignment
  • Support for both cloud and self-hosted models

GitLab Duo Self-Hosted

General Availability: February 2025

Enterprise AI deployment for organizations requiring complete data control.

Deployment Options:

  • On-premises (vLLM framework)
  • AWS Bedrock
  • Azure OpenAI
  • Air-gapped environments

Privacy Features:

  • No external API calls
  • Full control over request/response logs
  • On-premises model hosting
  • Complete data sovereignty

Target Industries:

  • Public sector
  • Financial services
  • Healthcare
  • Automotive
  • Regulated industries

Performance Optimizations

Prompt Caching

Default Enabled

Improves Code Suggestions latency by avoiding re-processing of cached prompts and input data.

Model Updates

  • GitLab 18.1: Codestral via Fireworks
  • GitLab 18.2: Claude Sonnet 4 (current default)
  • Continuous improvements to accuracy and speed

Adoption Statistics (2025)

  • 1.5 million developers using GitLab AI tools
  • 30% faster releases with AI-powered CI/CD
  • 6-month trend analysis available in Duo Analytics dashboard

Getting Started

Enable Duo Features

  1. For Premium/Ultimate:

    • Features enabled by default in GitLab 18.0+
    • Configure at group or project level

  2. For Duo Pro:

    • Subscribe at $19/user/month
    • Enable in group settings

  3. For Self-Hosted:

    • Requires GitLab Duo Enterprise subscription
    • Deploy AI Gateway and models
    • Configure model providers

IDE Setup

  1. Install GitLab extension for your IDE
  2. Authenticate with GitLab account
  3. Enable Duo features in extension settings
  4. Start coding with AI assistance

Usage Analytics

GitLab Duo Analytics Dashboard (18.7+):

  • 6-month trend analysis
  • Code generation volume tracking
  • Agent Platform flow adoption
  • Team-wide usage metrics
  • Feature-specific analytics

Compliance and Governance

Data Usage

  • Code context sent to AI models
  • No training on customer code (by default)
  • Configurable data retention policies
  • Full audit logging

Security Controls

  • Group-level feature toggles
  • Project-level overrides
  • User permission management
  • API rate limiting

Roadmap

February 2026

Global event showcasing GitLab as intelligent orchestration platform for software teams and AI agents.

Ongoing Development

  • Enhanced agent capabilities
  • Additional pre-built flows
  • Improved model performance
  • Expanded language support
  • Deeper IDE integrations

Resources

Next Steps