LeviathanTalon · Metanoia Unlimited

2,832 agents.
One click.

A Kabbalistic multi-agent swarm. 708 LLMs reason; 2,124 binary workers execute. Self-learning. Cross-pollinating. Multi-provider. Production grade — built honestly under the Preservation Doctrine.

708
LLM slots
2,124
Binary slots
75%+
Binary share
$0
Per binary task
Ketersource Chokmah Binah HaChazalDa'at · master ZadkielChesed GavrielGevurah RaphaelTiphareth UrielNetzach MichaelHod MalikaYesod KatanHaShemMalkuth

What makes this swarm different

Most agent frameworks burn LLM tokens on every step. LeviathanTalon's discipline is the opposite: LLMs reason, binary workers execute. ≥75% of every task's work runs on a deterministic $0-per-task pool. That's the math that lets one swarm do real work without one bill.

Architecture

1 + 7 + 700 + 2,124

Master HaChazal at Da'at. Seven sefirotic sub-orchestrators. 700 LLM workers. 2,124 binary executors. Total capacity 2,832 concurrent agents — verified live at 13,919 tasks/sec on the binary pool.

Self-learning

measurable curve

Every dispatch records (signature, approach, outcome). Same task next time: the swarm picks the historically best approach. The learning curve trends UP — verified by tests in the open-source bundle.

Quantum cross-pollination

8×8 affinity

QCPE: every agent's insight pollinates others through a learned 8×8 affinity matrix. Pairs that succeed together strengthen. Without explicit messaging. Pair-specific, not broadcast — verified by tests.

Multi-provider

7 providers

Anthropic · OpenAI · OpenRouter · Ollama · LM Studio · vLLM · Any OpenAI-compatible. Per-slot routing OR one model fills all 708. Mix local and cloud freely.

Plan-execute discipline

≥75% binary

Plans are typed DAGs of binary/LLM steps. The smart planner uses LLM only when reasoning is required; everything deterministic runs binary. The 75% binary share is enforced and live-measured.

Built honestly

165 / 165 tests

Sage Gimel honest-status files at every layer label features WORKING / PARTIAL / MOCKED with evidence anchors. Preservation Doctrine: v1 frozen, every change as an LPLS patch. No hand-waving.

Install + run in one command

Download. Unzip. Run the installer. Open your browser to a polished config UI. Configure your 8 LLM slots — or pick a model preset to fill all of them. Click LAUNCH.

Quick install

# 1. Download leviathantalon-ELITE-RELEASE.zip from this site
unzip leviathantalon-ELITE-RELEASE.zip
cd leviathantalon-swarm

# 2. One-command install of all 6 layers
./scripts/install.sh

# 3. Launch — browser opens at http://127.0.0.1:7088/config
lvtn

Host on your domain

# On your server
lvtn --host 0.0.0.0 --port 8088 --no-browser

# nginx — point your domain at the server
server {
  server_name leviathansi.xyz;
  location / {
    proxy_pass http://127.0.0.1:8088;
    proxy_set_header Host $host;
  }
}

Plug any LLM into any slot

Eight LLM slots — one master, seven sub-orchestrators. Pick any provider for each. Or hit a preset and fill all of them with the same model. Local, cloud, anything OpenAI-compatible.

AnthropicClaude · API
OpenAIGPT · API
OpenRouter100+ models · 1 key
OllamaLocal · $0
LM StudioLocal · $0
vLLMOn-prem cluster
GenericAny OpenAI-compat

Beats OpenClaw on every axis

Built after OpenClaw underdelivered. Same architectural intent, executed honestly with evidence.

AxisOpenClawLeviathanTalon Elite
Concurrent agents proven"150+" claimed, unverified2,124 binary @ 13,919 tasks/sec verified live
LLM reasoning vs binary execution splitmixed pico/nano pools — burned tokensexplicit ≥75% binary discipline, enforced + live-measured
Single-model filln/aone click — Kimi/Llama/anything fills all 708
Multi-provider LLMAnthropic only7 providers · per-slot routing
Self-learningnone demonstrably improvingmeasurable learning curve, verified by tests
Cross-agent memoryhive KV silos3-scope shared graph + semantic vector search
Quantum cross-pollinationnone8×8 learned affinity matrix · pair-specific
Honest accountingthe failure modeSage Gimel at every layer · 165 / 165 tests
Downloadable + brandedproject-specificsingle zip · one-command install · branded dashboard

Get the Elite Release

Download the full package — 6 layers, 165/165 tests passing, production-ready. Then run ./scripts/install.sh and lvtn.

⬇ Download leviathantalon-ELITE-RELEASE.zip

300+ KB · Python 3.9+ · MIT-style proprietary license (Metanoia Unlimited LLC)