The zero-setup install is the part people notice first. These are the parts they keep it for.
Both of them fail quietly in a bare terminal. Both are handled up front here.
Clootee checks connectivity before you hit that wall, and gives two concrete ways out: turn on your VPN and re-check, or switch to a reachable provider in one click. Only providers that actually respond right now are offered — a list you can trust beats a list that is merely complete.
Stock Claude Code must be signed in, or your messages get no reply and no error. Clootee moves sign-in into the UI: click the button, open the link, authorize, paste the code back. It works even when the server is remote, because the link opens in your browser.
A coding agent is not a chat window — a real task takes minutes, sometimes an hour. Clootee treats that as the normal case.
When an agent edits your files, “it worked” is not enough information. Clootee shows the whole trace:
The worst state a long-running agent can be in is “looks fine, is dead”. Clootee makes each of those states explicit, with a reason and a next step:
Each one is reported as what happened plus what to do about it, rather than a frozen screen you have to interpret.
projects/claude_hub/
├── backend/src/
│ ├── logic_struct/ # Orchestration — who is called, in what order
│ ├── logic_realize/ # Implementation — how each step actually works
│ ├── helper/ # Pure, business-free utilities
│ └── server/ # HTTP + WebSocket routes
├── frontend/ # Static, no build step
└── out_end/ # Portable runtimes, downloaded on demand“What it does” and “how it does it” live in separate files. The orchestration reads top to bottom in one pass, and changing a detail cannot damage the architecture. The frontend has no build step at all — what you read in the repo is what runs in your browser.
Install paths branch per operating system from one codebase, which is why the same zip serves Windows, macOS and Linux.
One zip, three platforms, Apache 2.0. No account, no telemetry.
Download Clootee v2.1.1