GitLab has launched the public beta of its Duo Agent Platform, an AI-driven orchestration layer built to enhance DevSecOps workflows by enabling asynchronous collaboration between human developers and intelligent agents.
Introduced on July 17, this new platform is designed to give AI agents meaningful awareness of a project’s context—its structure, history, and standards—so that they can contribute in ways that align with organizational goals. By embedding intelligence directly into DevSecOps pipelines, GitLab aims to streamline everything from writing code to reviewing and deploying it.
Key Features of the Duo Agent Platform
The beta release includes two initial AI agents:
- A chat-based assistant capable of handling general development queries, analyzing code changes, and interpreting issues
- A software development agent that performs hands-on tasks like writing code in virtual environments and submitting merge requests
These agents are now available to GitLab Premium and Ultimate users, both on GitLab.com and self-managed setups. Integration with IDEs such as Visual Studio Code and JetBrains is already in place via dedicated GitLab plugins, with more environments expected to be supported later.
Expanding Agent Capabilities
GitLab is actively developing specialized agents for other phases of the development lifecycle, including:
- Product planning
- Automated testing
- Code review
- Security scanning
- Deployment workflows
- Research and platform engineering
Each agent is intended to take on focused responsibilities, contributing to GitLab’s vision of modular, AI-supported DevSecOps.
Introducing Agent Flows
One of the more advanced features being developed is agent flows—automated workflows composed of multiple cooperating AI agents. For example, a software development flow (currently in beta) can coordinate agents to plan features, write and test code, and even respond to GitHub issues or merge requests, all based on deep contextual understanding of the codebase.
Integration with External Systems
To support complex environments, the platform also supports connections to external services using Model Context Protocol (MCP). This allows agents to gather context or perform actions in third-party systems such as Jira, ServiceNow, and ZenDesk. GitLab’s own MCP server is expected to be released soon to make such integrations more seamless.
Looking Ahead
The Duo Agent Platform is set for general availability by the end of the year. With its focus on intelligent automation and deeper contextual collaboration between humans and AI, GitLab is positioning this platform as a future cornerstone of modern software development.
