Hi Product Hunt 👋

I built KOLONİ because I kept juggling Claude Code, Codex and Copilot CLI in separate terminals, copying context between them and then reading diffs to see who changed what.

What it does:

• One chat with a live canvas. A lead agent works on your request or hands parts to roles (29 built in, plus your own), each on the CLI or model you choose.

• Each sub-agent works in its own copy of your project, so nothing touches your folder until you say so. Its changes come back on a card with the diff and your test command's result.

• Three levels: ask before changes, auto-edit files, or full auto. Stop keeps half-done work behind a card.

• It also runs a local llama.cpp server or any OpenAI-compatible API, and shows tokens, credits and plan windows per run as the CLIs report them.

What it isn't: it never signs in for you, and each CLI still sends your code to its own vendor as it always did. KOLONİ itself listens on 127.0.0.1 only and sends no telemetry.

It's 0.1.0, an alpha, open source under AGPL-3.0. I'd love to hear where it breaks for you, and which CLI or model you'd want next.