Skip to content

Why KOLONİ

An agent CLI on its own works in one terminal, one session at a time. KOLONİ puts the agents you already use, and the models you choose, in one place on your own machine, and shows you what they really did.

What it adds

  • Many agents, one chat. The Orchestrator hands parts of a job to other roles (Coder, Reviewer, Tester and more), each in its own working copy, so your folder changes only when you take the work in.
  • Any model you have. Claude Code, Codex, GitHub Copilot CLI, Qoder and agy on your own plans, an API key, or a model server on your own machine, side by side. Each agent shows the provider and model it really ran on.
  • Evidence, not a story. The canvas draws every turn from the record: the agents, their tools, commands and the files they changed, with real diffs. A value KOLONİ does not know is shown as unknown.
  • You decide. Before anything risky, an agent asks you on a card. The permission level sets how much runs without asking.
  • Memory with sources. Facts rest on lines of files and are checked again each time; a fact whose source changed is not used until you look.
  • Work that goes on. Teams follow recipes, triggers start work by themselves, and jobs run in copies while you are away.
  • Local. KOLONİ listens on 127.0.0.1 only, keeps its data in a folder on your machine, and sends no telemetry.

What it is not

  • Not finished: 0.1.0 is the first release, an alpha, so expect rough edges. See Known limits.
  • Not a sandbox by itself. Below Full auto your cards are the wall; for more, run commands in Docker. See Security model.