Unlock seamless collaboration between developers and AI! GitLab’s new platform automates tasks, boosting efficiency. #GitLabAI #DevSecOps #AIagents
Explanation video
Meet Your New AI Teammates: GitLab’s Big Update Explained!
Hi everyone, John here! Welcome back to the blog where we break down the latest in AI without the confusing jargon. As always, my wonderful assistant Lila is here to help keep me honest and make sure everything is crystal clear.
Today, we’re diving into some exciting news from a company called GitLab. They’ve just announced something that could change how programmers work by giving them a team of AI helpers. Let’s get into it!
So, What is GitLab Building?
GitLab has unveiled a new platform called the GitLab Duo Agent Platform. Right now, it’s in what’s called a “public beta.”
Lila: “Hold on, John. What does ‘public beta’ mean? Is the platform finished or not?”
That’s a fantastic question, Lila! Think of it like a new bakery having a “soft opening.” The ovens are on, they’re baking bread and cakes, and they’re inviting people in to taste everything. The customers’ feedback helps the bakers perfect their recipes before the grand opening. A public beta is similar—the software is working, and GitLab is letting its customers use it to find any little issues and gather feedback before its official, widespread release.
The main goal of this platform is to create a space where human developers and smart AI “agents” can work together seamlessly. GitLab calls this a “devsecops orchestration platform.”
Lila: “Whoa, that’s a mouthful! ‘Devsecops orchestration platform’? Can you explain that in plain English, please?”
Of course! Let’s break it down.
- DevSecOps: Imagine you’re building a state-of-the-art treehouse. ‘Dev’ (Development) is the part where you design and build it. ‘Sec’ (Security) is making sure the ladder is strong, the walls are sturdy, and there’s a railing on the balcony. ‘Ops’ (Operations) is about maintaining it, making sure it withstands the weather, and maybe adding new features later. DevSecOps is simply the idea of thinking about all three—building, safety, and maintenance—together from the very start.
- Orchestration Platform: This just means it’s a system that manages and coordinates all the different helpers (in this case, the AI agents) so they work together like a well-rehearsed orchestra.
So, in simple terms, the GitLab Duo Agent Platform is a system for managing a team of AI helpers that assist with building, securing, and running software, all at the same time.
What Can These AI “Agents” Actually Do?
This is the really cool part. GitLab isn’t just creating one giant, all-knowing AI. Instead, they’re building a team of specialized agents, each good at a specific job. The platform gives these agents all the necessary information about a project—its history, its code, its goals—so they can make smart decisions.
During the beta phase, two main agents are available:
- The Chat Agent: Think of this one as a helpful team assistant. It can perform general tasks like reading through project descriptions or looking at recent code changes to answer questions.
- The Software Development Agent: This agent is more hands-on. A developer can give it a task, and the agent can actually write the necessary code changes. It then packages these changes up neatly and submits them for a human team member to review and approve. It’s like having an AI junior developer on your team!
And this is just the beginning! GitLab has announced they are working on a whole crew of other specialized agents for the future, including:
- An agent for Product Planning
- An agent for Software Testing
- An agent for Code Reviewing (to help check a human’s work)
- An agent for Platform Engineering
- An agent for Security Analysis (to find vulnerabilities)
- An agent for Deployment Engineering (to help release the software)
- An agent for Deep Research
Teamwork Makes the Dream Work: Introducing “Agent Flows”
Having individual AI helpers is great, but what if they could work together on a big task without a human needing to micromanage them? That’s where “Agent Flows” come in.
An Agent Flow is a complex workflow where multiple AI agents team up and run on their own to complete a larger goal. For example, GitLab is testing a “software development flow.” In this flow, a developer might describe a new feature they want to add. The system then orchestrates multiple agents to get it done: one agent might plan the steps, another writes the code, and a third agent could test that new code to make sure it works correctly. All of this happens autonomously, with the agents using the project’s history and documentation for context.
Connecting to the Outside World
GitLab understands that companies use many different software tools to get work done. To handle this, the Agent Platform is being built to connect to other systems. The article mentions it can connect to “MCP (Model Context Protocol) servers.”
Lila: “Okay, you almost lost me there, John. What on earth is an ‘MCP server’?”
Haha, you caught me! It sounds complicated, but the idea is simple. Imagine our AI agents are chefs who only speak English. But the company they work for keeps its task lists in a tool like Jira (which we can pretend speaks French) and its support tickets in a tool like ZenDesk (which speaks German). An MCP server acts like a universal translator. It allows the English-speaking AI agents to understand the information from the French and German systems. This gives the agents more “context” or background information, so they can do their jobs much more effectively.
How Can People Try It?
As we mentioned, the platform is in a public beta. It’s currently available to customers who have GitLab’s Premium and Ultimate plans. Developers can interact with these agents directly from within their favorite coding software, which are often called IDEs.
Lila: “An ‘IDE’? Is that like a special type of AI?”
Good guess, but not quite! IDE stands for Integrated Development Environment. The easiest way to think of it is as a programmer’s ultimate toolbox. It’s a special application on their computer that has a powerful text editor for writing code, tools for testing it, and features for fixing bugs, all in one place. The new GitLab agents will work inside popular IDEs like Visual Studio Code and those made by JetBrains.
For everyone else, GitLab is aiming to have the platform fully launched and generally available by the end of this year.
A Few Final Thoughts
John’s Take: To me, this feels like a really practical and powerful step for AI in software development. Instead of a single “magic” AI that replaces people, this is about creating a team of AI specialists that augment human skills. It shifts the developer’s role from just writing every line of code to becoming a leader who guides a team of capable AI assistants. It’s an exciting glimpse into the future of building things.
Lila’s Take: I agree! Hearing it’s a “team” of AIs with different jobs makes it so much easier to understand. It’s not this one scary, super-smart thing doing everything. It’s more like how a real company works, with different experts for different tasks. It makes the whole concept of AI helping us feel a lot more friendly and collaborative!
This article is based on the following original source, summarized from the author’s perspective:
GitLab introduces AI agent-enabled devsecops
platform