Skip to content

Quick start

From install to your first approved change, in five steps.

1. Install and start

Follow Install for your system. When KOLONİ starts, its window prints one line:

KOLONİ is open at http://127.0.0.1:8790/#key=…

Started from a terminal, KOLONİ also opens this page in your browser. If it does not, open the link from the window. Keep the window open while you use KOLONİ; Ctrl+C in it stops KOLONİ.

On first open, the canvas shows First steps: a provider, a folder, and your first message. The step you are on has a button that takes you where you do it.

2. Add a model

Press Add a provider (in First steps, or on the Models page). Fill in:

  • Name: any name you like.
  • Kind: OpenAI-compatible server, Anthropic API, or a CLI: Claude Code (CLI), Codex (CLI), GitHub Copilot CLI, Qoder (CLI) or agy (CLI).
  • For a server: its Address (a llama.cpp server is usually http://127.0.0.1:8080/v1) and an API key if it needs one.
  • For the Anthropic API: your API key and a Default model.
  • For a CLI: nothing more, as long as the CLI is installed and you signed in to it in your own terminal. KOLONİ never signs in for you. If the CLI is missing or not signed in, its card shows the command to install it or sign in.

Press Save, then Test to check that it answers.

The first provider on the Models page is the default. KOLONİ makes one called "Local" at the first start, for a model server on this machine. If you do not run one, open your new provider and press Make default.

If a CLI was installed when KOLONİ first started, you may find it on the Models page already.

3. Choose a folder

Press the folder button at the top left and choose Open a folder…. Pick the project your agents should work in. To only talk, without files or commands, choose No folder · a plain chat.

4. Write your first message

Next to the message box, check the permission level. Keep Ask before changes for now: each file change and each command will wait for you. See Permission levels.

Write what you want done, for example "Add a README that explains how to run this project", and send it.

The chat shows the Orchestrator's answer and a row for each agent it opened. The canvas draws the turn: each agent as a box with its tools, commands and files.

5. Approve a card

When an agent wants to change a file, a card shows the change as a diff. Press Approve to write it, or Reject to turn it down. After it is written you can still Revert it.

A command asks with "Run this command?":

  • Run once runs it this time.
  • Always run … runs it, and keeps a rule so the same kind of command runs without a card in this chat.
  • Reject turns it down.

While something waits for you, the project's tab above the canvas turns orange, and Ready on the rail counts it, so you can do something else meanwhile.

Next