For the last 40 years or so, software development tooling has centered around the IDE — the integrated development environment. Borland is widely credited with bringing the IDE to the masses. The joining of the editor, compiler, and debugger into a single entity revolutionized the software development industry. The popularization of the IDE through the likes of my beloved Turbo Pascal in the mid-1980s marked a huge turning point for coding. But the IDE’s reign is over. With the unbelievably rapid rise of agentic coding, the IDE has become an afterthought, a tool that developers are finding they use less and less. Instead, developers are spending time managing agents that are actually writing the code.
Things are moving pretty fast, changing monthly if not weekly. It was only a few months ago that I was building things inside my IDE with an agent’s help. But before long, I had four or five console windows open, working on different projects all at the same time as the agents ground away at different things. This was barely manageable when working on separate repositories and projects at the same time, but when I wanted to work on different features inside the same repository, it got a bit hairy.
Clearly, we developers have a need for a next-generation tool to manage all of this.
The first step was recognizing a little-known and underutilized feature of Git — worktrees. Git worktrees allow you to check out multiple branches of the same repository into different directories from a single Git database. A longtime power-user technique used only by a few, worktrees are perfectly suited for the new world of agentic coding.
Typically, a developer would work on one ticket at a time, and a simple Git checkout would do the trick. That assumes one actor — the developer — working on the codebase. But with coding agents, the notion of working on three Jira tickets at the very same time is no longer crazy, and worktrees enable that. They give each “developer” their own branch and directory. It’s easy isolation without the overhead of cloning and forking separate instances of repositories.
So this old Git feature that has been lying around basically unused has suddenly become the key to agentic coding. But of course, the overhead of worktrees is challenging in and of itself. If only there were a way to manage all of that in one place.
And that is where ADEs come in — agentic development environments. A new kind of development environment that arose with agentic AI, an ADE coordinates and manages all of the things that need to happen when multiple coding agents work on multiple issues in multiple branches in one repository. As Matt Johnston, CEO of GitKraken, maker of the Kepler ADE, put it, “The IDE was built for the age of one human typing. The ADE is built for the age of humans orchestrating fleets of agents.”
ADEs will do all the dirty work of opening, running, and closing the worktrees in Git. Having five console windows open and managing all of that is not something a human is very good at. But an AI-driven agentic development environment sure is.
The days of single-threaded development are over. With an ADE, developers can multitask with ease and track many tasks at once. Without having to worry about the logistics and overhead of managing multiple agents working in multiple worktrees, developers can focus on directing the agents and ensuring that they do the right work for the task at hand.
Admit it, you don’t really use the IDE that much anymore. And you aren’t going to miss it.