GitKraken Reveals Code Flows to Help Teams Navigate the Age of AI
As AI adoption accelerates and the nature of code changes fundamentally, GitKraken has introduced Code Flow, a framework to define how work flows between developers, code agents, repositories, reviews, pull requests, and finally the production environment. This, because AI has become more widespread removes developers from writing all the lines of code themselves and refers to the role of controlling, coordinating, and updating the result of many AI agents at the same time..
This new way of working creates a complex environment. Historically, development tools were designed to track code written by humans. Now, developers have to manage a much larger volume of work, including multiple branches, pull requests, and commits, all generated by agents across multiple clusters. As GitKraken CEO Matt Johnston noted, the difficulty is no longer about generating code—it’s about turning all that agent work into high-quality, finished software..
“Eleven years of watching code run on real teams has taught us where things go wrong,” he said. “But what we see now is different in scale. The AI has multiplied the volume of code flowing through those same points of failure. An update that has never been taken is now three updates. The collision that blocks it is now five. A branch that went too far is so big that the developer you opened is afraid to touch it, which used to happen from time to time. Now it happens to every athlete.”
This is where the concept of code flow comes into play. Code Flow is designed to describe the entire life cycle of software development in an agent-heavy environment. It tracks how the work goes through all the stages: from initial planning and coding to revision, compilation, and finally, the production environment.
Who is Code Flow?
As Johnston explained to SD Times: “We think of three audiences: the coding agent, the developer, and then the engineering lead. And the way we think about it is that if everything goes well, if the code flows from the system to the master at the speed of the agent’s development, the agents work for the developers, and the developers work for the engineering lead.” He said that agents can be thought of as the front-end production force, whether for planning, or coding, or updating, but ultimately they “report” to the developers who plan that work, sign it, or intervene in the work if necessary. The engineer then reports to the manager, who reports to the VP, who wants code flow because he wants to work 50% faster. “It’s all the things around code generation that create tension from, say, Claude Code has generated some code, what are all the things that need to be true in order for me to send this to production in a safe and reliable way?”
The framework focuses on the key elements of visibility, governance, and integration. Because agents are creating a huge explosion in code speed and volume, teams need a better way to understand where work is getting stuck and how individual agents are performing. Code Flow is a discipline for maintaining standards and keeping context clear as AI-driven development becomes the norm.
Introducing Kepler
GitKraken recently launched new products where code can flow. The first is Kepler, a purpose-built agent development environment (ADE), which resides on the Windows, Mac and Linux desktop. The company also introduced GitKraken Desktop 12 with a new agent mode, and GitLens 18 – a popular IDE extension to VS Code Cursor and anti-gravity ecosystem – which adds agent capabilities.
Johnston said Kepler is part of what the company calls the developer experience with Kepler, GitKraken Desktop and GitLens. “GitKraken Desktop and GitLens are designed to manage Git and help resolve conflicts, and have greater code visibility,” he said. “They’re not designed to say, ‘How do I send 20 agents at once, turn them into 200 agents?’
From program to main
The entire system around code generation was not designed for today’s volume or speed, and includes everything from code review to conflict resolution to code context, producing the right type and size committed to the right level of definition. Johnston went on to explain how customers will use the Code Flow framework. “You can work in a kind of traditional Git mode, that’s GitKraken Desktop. If you’re working in code mode, that’s in your IDE, that’s GitLens. And if you’re in first agent mode, that’s Kepler. And what we’re getting early on from our preview users who have been using Kepler for the past few weeks is that they say, ‘It’s not really my one job or’ usually it’s not the first or another day. Kepler, I spin my agents 12, I do ux, y, and z, then I can see that I need to re-enter some code through GitLens, with Cursor into my IDE, and before I compile, I realized that I have a big kind of bad conflict organized in one program and one subscription, because they are more ways of the developer’s work than three different products.”
