Hacker / defender · many terminals · is it Code-Hub? · 21 Aug 2026
No. Code-Hub is not the AI sitting in those windows. ParviClaw Core is a factory that opens terminals (about 64 live ones by default — that is the “65+” number). It does not think. Each window runs the command the hacker or defender asked for. If they ask for nothing, the window just waits. When they ask a model through a separate door, the brains are LLM-Hub (Claude, Grok, Qwen, and the rest). Code-Hub is the board that sees the work, plus a small picker that chooses which model LLM-Hub should use. It does not type code in the 64 panes.
Code-Hub = board + picker. Not 64 copies of Code-Hub coding.
| Thing | Is it coding in the 64 windows? |
|---|---|
| ParviClaw Core | Opens / watches / records. Does not code. |
| Command they named (Grok, Claude, script) | Yes — that is the coder in that window. |
| LLM-Hub | Yes for the separate “ask a model” door. Not auto-filled into every pane. |
| Code-Hub | No. Board + picker. Not the pane brain. |
Older notes said Core had no model door at all. Now it has one: it forwards to LLM-Hub. That did not put Code-Hub inside the terminals. We did not duplicate Grok-Build either. Two doors, same Core.
Only you: if you want those 64 windows to actually run a coding AI (Grok, Claude, …) instead of empty waits, say so. Today that is not the default. Core will not invent Code-Hub as the coder.