One model on all 708 LLM slots
Pick a single provider + model. Hit FILL. Every slot — master, 7 sub-orchestrators, all 700 LLM workers — is backed by your chosen model. Mix providers later if you want; this gets you running fastest.
Fill all 8 slots
Examples: Kimi K2.6 1T via OpenRouter, Llama 3.3 70B via Ollama, GPT-4o-mini direct from OpenAI. With per-sub concurrency at 100, all 708 LLM positions run in parallel on the same model.
Kimi K2.6 (via OpenRouter)
Llama 3.3 70B (Ollama local)
GPT-4o-mini (OpenAI)
Claude Haiku 4.5 (Anthropic)
Qwen 2.5 72B (LM Studio)
Adjustable scale
Tune concurrency per tier. The LLM slider sets the per-sub LLM concurrency (1..100; total LLM tier = 1 master + 7 subs + 7 × this value). The binary slider sets the unified ZeroClaw pool (0..2,124).
LLM per sub
1..100 — 7 subs × this = LLM workers
1..100 — 7 subs × this = LLM workers
100
Binary pool
0..2,124 — unified ZeroClaw concurrency
0..2,124 — unified ZeroClaw concurrency
2,124
Total concurrent capacity
2,832 agents
Or configure each slot independently
Mix and match: Anthropic for HaChazal, local Ollama for Raphael, vLLM for Uriel — anything works.