KabulEvery coding agent at its best, side by side

Run Claude, GPT, Gemini, Qwen and local models in parallel, each in the harness made for it, each on its own git worktree. Switch models without losing the conversation, bring a friend into the session and keep going from your phone.

Quick start

Install

Português

🧭

The right harness for every LLM

Claude in Claude Code, GPT in Codex, Gemini in Antigravity or Gemini CLI, Qwen in Qwen Code, everything else through OpenCode or Goose, over ACP. Local models are routed by family.

🔀

Switch LLM or account mid-conversation

Same chat, same branch, same worktree. Kabul hands off the latest messages, the tools used and the diff to the new agent, and checkpoints keep going.

🤝

With a friend, and from your phone

Invite an editor or an observer into a workspace, share one queue and one terminal, and follow every session from an iOS or Android app served by your own computer.

🌳

One workspace per task

Each task is a git worktree on its own branch, with reserved ports, setup and run scripts, a terminal and any number of chats.

✅

Review and ship

Split diff with comments for the agent, checks with logs, one button for commit, PR, fix, merge and archive. A checkpoint before every message.

📈

Built for scale

100 workspaces and 30 agents at once without the interface slowing down, with a concurrency limit, usage tracking and budgets measured in CI.

Kabul: parallel workspaces, a diff with a line comment and the chat waiting for an approval

Create a workspace, let Claude work, switch to Codex in the same chat, approve and open the PR