GitLab has taken a step forward in the integration of artificial intelligence in the day-to-day of software development with the recent release of the public beta of its Duo Agent PlatformThis move introduces a renewed approach for development teams seeking to combine the efficiency of AI agents with the strategic vision of human developers, facilitating collaboration even when teams work in a distributed manner or on different schedules.
GitLab's bet is to offer a DevSecOps orchestration platform that allows artificial intelligence agents to access a complete context of the project. It is intended that decisions made by AI are aligned with the standards and needs of each organization, increasing trust and reducing potential communication failures in complex processes.
Key features in the public beta
Among the most relevant developments of the Duo Agent Platform, highlight multi-agent workflows capable of gather the entire context of a projectAgents can answer questions directly with developers and execute precise changes to repositories, leveraging GitLab's structure, code history, issues, and merge requests to perform complex tasks autonomously.
GitLab's Agentic Duo chat is another of the most striking features. Available in both major IDEs and the GitLab web interface, this chat transforms the classic question-and-answer window into a true development companion. Thanks to it, programmers can delegate tasks using commands like /explain, /tests and /include, facilitating the transfer of context from files or incidents and customizing agent behavior with your own rules.
Expanded integration and technical support
With the new beta, GitLab extends the Support for the JetBrains family of IDEs (IntelliJ, PyCharm, GoLand, WebStorm), adding these widely used tools in the industry to the integration already available with Visual Studio Code. Current users can automatically access Duo Agentic chat, while new users can easily install it from the corresponding marketplace.
Another notable technical aspect is the implementation of the Model Context Protocol (MCP), allowing GitLab agents to connect to both local and remote MCP servers. This expands the possibilities for interaction outside of the GitLab environment, as agents can communicate with third-party systems and services that expose their data via this protocol.
Seamless collaboration and personalization
The proposed work model allow that AI agents act as additional members of the DevOps team, helping with both individual tasks and processes that affect the entire group of engineers. Developers can set custom guidelines, creating flows and rules tailored to the needs of the project and ensuring that human-AI collaboration is flexible and tight to the reality of each organization.
GitLab is also working on what it calls a “Agent Catalog”, a marketplace where organizations can create, customize, and share agents or agent flows, thus enriching the ecosystem and facilitating the adoption of solutions tailored to specific problems.
Commitment to updates and vision for the future
The company has committed to publishing monthly improvements on the platform, with iterative releases under version 18.x and a focus on achieving general availability before the end of the year. The goal is to expand the capabilities of the Duo Agent Platform, as well as refine the user experience and adapt to the new demands of the technology industry.
Initial reviews from companies and industry experts are positive. The following stand out, in particular: the ability of agents to understand the full context of the code and business needs, allowing developers to focus on higher-value, creative tasks. This GitLab approach represents an evolution for DevSecOps work, transforming it into much more intelligent and connected processes.
This release marks a significant advance in the integration of artificial intelligence into development processes, enabling the automation of repetitive tasks, minimizing errors, and boosting team efficiency. Intelligent orchestration and collaboration between humans and AI agents will represent the next major leap in the productivity and quality of deployed applications.