Model
defaultVoice
Prompt
Voice commands (say "computer …")
reflective (default): self-corrects on failure.
react: thought → act → observe loop.
planner: writes a plan first, then executes.
Changing it only affects the next run.
Response
Select tools to enable for Agent runs, or execute one directly. Toggle Examples on, then click a tool name to fill the runner with a ready-made call.
All registered sites. Start/stop by port, stage releases with git worktrees (dev → beta → prod), and mirror the checked sites to D:\Backups\Kucatoo.
| Health | Instance | Tier | URL | Custom domain | Status | Actions |
|---|
Recall from tiered memory (STM/MTM/LTM), or remember a new fact.
Browse installed skills and execute one with parameters. Toggle Examples on, then click a skill name to fill the runner with a ready-made call.
Browse the active workspace. Select files with the checkboxes, then use the toolbar.
Cost per project (each subdirectory of the workspaces root is a project). Click a project to see its queries; click a query row for the full prompt, response, and settings.
Provider speech (MiniMax / Qwen) alongside the browser default. Set a budget to cap spend; every call is priced and logged to Usage.
Wake-word personas: address the app by any persona name below instead of "Computer" and spoken replies & voice chat switch to that persona's voice. Saying "Computer" returns to the defaults above. Names are editable — use one lowercase word, 3+ letters (distinctive multi-syllable names like "Xander" recognize better than short common ones). Columns: browser fallback gender · MiniMax voice (spoken replies) · Omni voice (STS voice chat; beyond Chelsie/Serena/Cherry/Ethan needs Omni-Flash, not Turbo). Changes apply immediately; "Save settings" keeps them server-side.
Clone a custom voice from a short audio sample (MiniMax 10s–5min, Qwen 10–20s). The cloned voice then appears in the Voice picker above.
Live harness status: capabilities, provider health, cost.
Guest access links
Shareable trial links: guest sandbox dir, cheapest models, no video/vision, capped queries/tokens, expiring.
Every completed run is recorded automatically. Click an entry to reload its prompt and response.
Your fleet: the worker CLIs Kucatoo delegates to, and the instances you run. Every instance is one provider-specific CLI — Kucatoo calls them and coordinates the work.
Workers you can rent
A slot is one worker position. The CLI decides the provider; Kucatoo-Code is the one where you choose the provider yourself.
Buy a slot
Pick a size and a worker and it is added to your plan — billed per slot per month, on the same invoice, cancel any time. The machine starts when the payment clears.
Your slots
What you are paying for, read from your subscription — so this list always matches your invoice.
| Worker | Size | Price | Actions |
|---|
Run work on a slot
Send a prompt to one of your own slots. It runs at that slot's size, on your provider key from Settings.
Slot machines
The runtime view of those slots: which model each machine is set to, and any upgrade offer when one keeps running out of memory.
| CLI | Provider | Model | Size | Price/mo | Capacity failures | Actions |
|---|
Your instances
The virtual machines you are paying for, newest first.
| Status | Instance | Tier | Memory | Subdomain | Client |
|---|