GitKraken shipped Kepler on June 15—a standalone Agentic Development Environment (ADE) built for a job description that did not exist two years ago: directing multiple AI coding agents across multiple repositories and getting the output merged. Kepler does not replace your IDE or your agent. It sits on top, connecting issue trackers, Git providers, and whichever agent runtime you already use into one delivery surface.
ADE vs IDE
GitKraken draws a deliberate product line. An IDE was built for one developer typing code in one repo. An ADE is built for one developer running several agents in parallel—planning work, launching sessions, reviewing diffs, resolving conflicts, and pushing to merge without living in five terminal tabs. Kepler is agent-agnostic at launch: Claude Code, Codex CLI, Copilot CLI, Cursor, and OpenCode all connect as agent sessions inside Tasks.
GitKraken Desktop remains the single-repo, single-agent tool. Kepler targets the next step when parallel agents become the default and branch management becomes the bottleneck.
Tasks as the Unit of Work
A Task is Kepler's core abstraction. It spans one or more repos, holds isolated Git worktrees (one per repo per Task), tracks agent sessions, and carries shared context across the effort. A database migration touching API, worker, and frontend repos is one Task—not three disconnected branches you reconcile manually at the end.
Launch flow: pick repos from the Task Launcher, name the Task, optionally seed a prompt, choose agent/model/mode/effort, and click Launch. Kepler clones missing repos, creates worktrees, and starts the agent session with repo state wired in automatically.
Delivery Engine Features
- Multi-Agent Oversight. Kanban and list views show every agent session across every repo—what is running, what needs human input, what is ready to merge.
- Conflict Resolver and Auto-Rebasing. Parallel agents create parallel collisions; Kepler surfaces conflicts as first-class operations and keeps branches current so stale-branch rebases stop eating your afternoon.
- Branch Intelligence. Live branch state, conflict risk, and merge readiness appear in the UI—not inferred from a terminal prompt.
- AI Sync (experimental). Optional agent tools for rebase/merge with automatic conflict resolution, each operation rollback-able from Settings → Features.
Agent Integrations
Kepler launches your existing agent binary; it does not ship a proprietary model. Each supported agent runs as a session inside a Task with Task context and repo access passed at start. Agent choice is fixed for the life of a session but model, mode, and effort can be overridden per message where the underlying agent supports it.
Remote environments—including SSH and WSL—are supported. A local server mode lets you control Kepler from another device when your agents run on a beefier machine.
Where Kepler Fits the Stack
Kepler closes the gap between "code generated" and "code merged"—the same problem Cursor 3 attacks from inside the IDE and JetBrains Central approaches from the JetBrains ecosystem. Kepler is Git-native and agent-neutral: useful if your team already mixes Claude Code in one repo and Copilot CLI in another, or if you are scaling from two parallel agents to twenty and need a control plane.
Kepler is free during a limited preview on Windows, Mac, and Linux. If you are already shepherding branches for multiple agents, connect your existing setup and treat the first Task as a trial: one cross-repo change, two agents, one merge—see whether the oversight surface saves more time than it adds.