[2/2] GitLab Act 2How much it matches GitHub, Atlassian and Agent IDE (Category AI4SDLC)
In past I took down the GitLab Act. 2 as an attempt to rebuild the DevSecOps platform for agent development: scale Git for machine use, implement the orchestration of the entire life cycle, collect a context graph, embed governance and implement a hybrid model of human/agent/autonomous work. And in this post, I wanted to compare it to what other platforms are doing and how much the course is aligned.
TLDR; The course matches strongly, but the difference is where each player tries to become a control plane for agent development.
Then let’s look at the other players.
1️⃣ GitHub GitHub comes very close, but with different optics. Copilot coding agent already works as an asynchronous member of GitHub flow: he can be assigned an issue, he works in the environment based on GitHub Actions, flies commits in draft PR, and the person reviews the result. And Agent HQ explicitly describes GitHub as mission control for different agents: Copilot, OpenAI Codex, Claude, Google, Cognition and others. GitHub’s bet is to leave the usual primitives: issues, pull requests, Actions, branch protections, audit, but make agents native inside the process. I read more in a separate article in my blog tellmeabout.tech
2️⃣ Atlassian Atlassian looks at the same challenge through workflow and corporate context. Rovo Dev lives alongside Jira, Confluence, Bitbucket, and GitHub, and Atlassian’s strength is Teamwork Graph: connecting tasks, requirements, documentation, team knowledge, and code. If GitHub says “agents within the repository and PR,” Atlassian means “agents within the company’s workflow.” I read more in a separate article in my blog tellmeabout.tech
3Cursor and JetBrains play more IDE-first Cursor relies on background agents and remote environments where the agent works in a separate branch. JetBrains’ Junie is built into the IDE and draws on the familiar development environment, inspections, tests and project context. Their battlefield is the developer experience: where an engineer reads code, watches diff, and makes a decision.
4OpenAI/Codex and Devin-like agents go even more agent-first There, the main interface is not the repository, Jira or IDE, but the agent as a working unit: received the task, raised the environment, changed the code, checked, returned PR.
In general, you can see that all the major players see trends and try to grab them by the tail, but each in its own way. Here GitLab connects agency not only with coding assistant, but with Git, CI / CD, governance, data model, pricing and organizational structure. And in this model, we see the role of the engineer shifting. There is less value in having your hands do every action. It’s more about setting intention, gathering context, defining constraints, building validation, making architectural decisions, and keeping quality at a faster rate of change.
There is a great struggle for who will occupy the control plane niche. GitHub has a game around PR and repository. Atlassian has a game around tasks and organizational knowledge. Cursor and JetBrains have an IDE game. OpenAI/Codex has a game around the agent as a new work interface.
And GitLab is trying to take the broadest place: control plane for the entire software lifecycle in enterprise. The bet is strong, but the risk is also high. In agent development, the big question is, will your engineering system withstand the speed that agents create?
#AI #AI4SDLC #Engineering #DevSecOps #Management #Architecture