GitLab Duo Agent Platform beta unlocks AI-human collaboration

GitLab has opened public beta access to its GitLab Duo Agent Platform, a DevSecOps orchestration platform enabling asynchronous collaboration between developers and AI agents.

Product details

The GitLab Duo Agent Platform introduces an orchestration layer designed to allow specialised AI agents and human developers to collaborate within software development projects.

By leveraging GitLab as the system of record, the platform delivers broad project context to AI agents, supporting informed decision-making in line with organisational standards.

The company has made the public beta available to Premium and Ultimate customers.

The initial set of features includes Software Development Flow - the first orchestrated multi-agent workflow that accumulates context, clarifies ambiguities with developers, and implements changes to codebases and repositories using project structures, codebase history, and supplementary context such as GitLab issues and merge requests.

Specialised agents and workflows

Specialised agents on the platform mirror established team roles, with capabilities to search, read, create, and modify existing artefacts across GitLab.

The platform also features agent Flows, which are structured, predetermined workflows that can coordinate multiple specialised agents to autonomously execute complex or multi-step tasks.

GitLab is planning an AI Catalogueueueueue in the future - this marketplace will allow organisations to create, customise, and share agents and agent flows among their teams and the wider GitLab ecosystem.

Interface and support

Users of the public beta have access to GitLab Duo Agentic Chat within development environments, both in IDEs and the GitLab Web UI.

According to GitLab, the chat experience has been transformed into an active development partner, supporting iterative feedback and chat history, as well as streamlined delegation using new slash commands such as /explain, /tests, and /include. These commands create a quick delegation language, and the /include feature allows for context injection from specific files, issues, merge requests, or dependencies.

Developers can also personalise agent behaviour using custom rules, specifying guidance tailored to individual or team preferences through natural language instructions.

In addition to integration with Visual Studio Code, support has been extended to JetBrains IDEs such as IntelliJ, PyCharm, GoLand, and WebStorm.

The platform also introduces Model Context Protocol (MCP) Client Support, which enables GitLab Duo Agentic Chat to connect to remote and local MCP servers.

This allows agents to communicate with systems beyond GitLab, provided those systems are accessible via MCP, expanding the practical application of the platform's capabilities.

Future releases

GitLab stated that the scope and quality of the Duo Agent Platform will be expanded through subsequent 18.x releases, with a general availability target by the end of the year.

Industry perspectives

GitLab's own leadership and industry observers offered perspectives on the platform's beta release.

"GitLab Duo Agent Platform enhances our development workflow with AI that truly understands our codebase and our organisation," said Bal Kang, Engineering Platform Lead at NatWest.

"Having GitLab Duo AI agents embedded in our system of record for code, tests, CI/CD, and the entire software development lifecycle boosts productivity, velocity, and efficiency. The agents have become true collaborators to our teams, and their ability to understand intent, break down problems, and take action frees our developers to tackle the exciting, innovative work they love."

Rachel Stephens, Research Director at RedMonk, commented, "As software development workflows grow in complexity and organisations look to leverage AI, there's an increasing need for platforms that can integrate AI capabilities without adding to existing disjointed toolchains."

"As a DevSecOps platform, GitLab is already positioned to help developers collaborate both synchronously and asynchronously. Now the GitLab Duo Agent Platform intends to take this a step further, helping developers also integrate AI agents into their workflows."

Bill Staples, Chief Executive Officer at GitLab, added, "Today marks a pivotal moment in software development as we introduce the public beta of the GitLab Duo Agent Platform, the first DevSecOps orchestration platform designed to unlock asynchronous collaboration between developers and AI agents."

"GitLab Duo Agent Platform isn't just another AI tool; it's a fundamental reimagining of software development from isolated, linear processes into dynamic, intelligent collaboration."

"By leveraging GitLab's unique position as the system of record for the entire software development lifecycle, we're providing AI agents with unprecedented context and capabilities. This enables our customers to work with AI agents that have comprehensive context about their codebase, their workflows, and their organisational goals to help boost productivity, velocity, and efficiency."