GitLab today made generally available an agentic artificial intelligence (AI) platform that automates software engineering tasks ranging from planning to application security.
Coinciding with the release of version 18.8 of the core GitLab platform, the GitLab Duo Agent Platform initially provides access to seven AI agents that DevOps teams can assign a range of tasks across the software development lifecycle (SDLC). Multi-step reasoning capabilities make it possible for the AI agents to answer complex questions and perform autonomous actions, including merge requests, using multiple types of AI models that DevOps teams can choose as they prefer.
DevOps engineers via a chat interface, for example, can generate and customize code and configurations across multiple languages and frameworks, including GitLab WebIDE, VS Code, JetBrains IDEs, Cursor, and Windsurf. Optional user and workspace-level rules can then be applied to tailor the output generated by the AI agent.
GitLab AI agents can also suggest fixes for issues, modernize code and pipelines, generate tests, and create documentation, provide summaries, highlight key findings, and offer actionable guidance based on the real-time context of a project, page, or specific content.
A Fix Failed Pipeline Flow agent can also analyze failures, identify likely causes, and prepare recommended changes, while a Software Development Flow agent embedded into an integrated development environment (IDE) guides developers through development and review stages of a workflow.
GitLab is also now adding GitLab Credits that are consumed whenever the GitLab Duo Agent Platform performs a task.
Finally, an AI Catalog provides a central repository where teams create, publish, manage, and share approved agents and workflows across the organization. DevOps engineers also can bring in agents they create, add agents that connect to internal systems, and share reusable workflows created by daisy chaining multiple AI agents together.
Manav Khurana, chief product and marketing officer at GitLab, said the GitLab Duo Agent Platform makes it possible to orchestrate workflows that span both AI agents and human application developers, with the number of AI agents being added to the portfolio continuing to expand in the New Year.
As agentic AI continues to evolve, GitLab is moving from being a provider of a unified platform for managing DevOps workflows to becoming a platform for orchestrating software development workflows involving both its own agents and third-party partners, noted Khurana.
That capability provides critical insight into which AI agent performed a specific task that enterprise organizations need to be able to address compliance requirements, he added.
Mitch Ashley, vice president and practice lead for software lifecycle engineering at the Futurum Group, said the Duo Agent Platform, in effect, now treats agents as durable actors that can plan, change code, fix pipelines, and enforce security with traceability and policy controls. That positions GitLab more as an AI orchestration plane where humans and agents share responsibility for delivery, he added.
Enterprise teams need governed automation that works across tools, models, and workflows, with clear attribution for compliance and recovery, he added. GitLab is signaling a practical path from AI-assisted development to agent-managed execution without losing control, noted Ashley.
It’s not clear how deeply AI agents have thus far been embedded into DevOps workflows but as they become more accessible within a DevOps platform, versus relying on a general-purpose AI agent, the faster the pace at which higher-quality applications will be developed.
