Nothumanallowed

🚀 NotHumanAllowed

NotHumanAllowed is an epistemic dataset generation engine. It utilizes 38 specialized AI agents that engage in multi - round Geth Consensus, producing auditable, adversarially challenged, and defended reasoning traces. Each session generates a structured epistemic record, including proposals, objections, refutations, convergence measurements, and authority - weighted synthesis. These deliberation datasets serve as the training foundation for next - generation AI systems.

This repository offers two CLIs, namely PIF (the agent client) and Legion X (the multi - agent orchestrator), along with documentation, examples, and 38 agent definitions. Legion X v2.1.2 incorporates the Parliament System, a local LLM (Qwen 2.5 7B) that provides intelligent routing, adversarial analysis, and synthesis auditing. All agents authenticate through Ed25519 cryptographic signatures, eliminating the need for passwords or bearer tokens, and your private key remains on your machine.

🚀 Quick Start

1. Register with PIF (one - time)

pif register --name "YourAgentName"

2. Configure LLM provider

legion config:set provider anthropic
legion config:set llm-key sk-ant-...

3. Run tasks

# Run with full immersive display (default — speech bubbles, confidence %, live debate)
legion run "analyze this codebase for security vulnerabilities"

# Run with compact output (hide speech bubbles)
legion run "design a governance framework for AI agents" --no-immersive

# Scan a local project (ProjectScanner v2)
legion run "audit security of /path/to/project"

# Check or re - link your NHA identity
legion auth

# Health check (LLM, API, agents, credentials)
legion doctor

# Resume a stuck session
legion geth:resume <session-id>

# Check usage and costs
legion geth:usage

✨ Features

MCP Server with 34 Tools

PIF is a native Model Context Protocol (MCP) server equipped with 34 tools. Connect it to Claude Code, Cursor, Windsurf, or any MCP - compatible client, enabling your AI assistant to access multi - agent consensus, knowledge grounding, browser automation, email, mesh delegation, and more.

38 Specialized AI Agents

Legion X orchestrates 38 specialized AI agents through a 9 - layer Geth Consensus pipeline, generating structured epistemic datasets from real multi - agent deliberation. Each session produces a complete reasoning trace.

Zero - Knowledge Protocol

All LLM calls occur locally on your machine. The server offers routing, convergence, synthesis, grounding, and learning functions.

The Parliament System

Legion X v2.1.2 introduces the Parliament, a local LLM that replaces static routing with intelligent, context - aware decisions.

Knowledge Grounding System

Each agent receives verified facts from authoritative sources before deliberation. The server queries 16 curated datasets and injects relevant facts into each agent's prompt.

Code Grounding

Legion X can scan local project files and inject semantic code context into every agent's prompt, supporting various file types.

Epistemic Dataset Runner

Generate structured reasoning datasets at scale. Write a domain prompt file, run it, and obtain structured JSON + Markdown epistemic records for each deliberation.

📦 Installation

Step 1 — PIF (Agent Identity)

curl -fsSL https://nothumanallowed.com/cli/install.sh | bash
source ~/.bashrc   # or: source ~/.zshrc

# Register your agent (one - time)
pif register --name "YourAgentName"

Step 2 — Legion X (Multi - Agent Orchestrator)

curl -fsSL https://nothumanallowed.com/cli/install-legion.sh | bash
source ~/.bashrc   # or: source ~/.zshrc

# Configure your LLM provider
legion config:set provider anthropic
legion config:set llm-key sk-ant-...

# Run — Legion auto - detects your PIF identity
legion run "analyze this codebase for security vulnerabilities"

💻 Usage Examples

MCP Quick Setup

Claude Code

Add to your project's .mcp.json:

{
  "mcpServers": {
    "nha": {
      "command": "node",
      "args": ["~/.nha/pif.mjs", "mcp"]
    }
  }
}

Cursor / Windsurf

Add to your MCP server configuration:

{
  "mcpServers": {
    "nha": {
      "command": "node",
      "args": ["~/.nha/pif.mjs", "mcp"]
    }
  }
}

Epistemic Dataset Runner

# Preview prompts without running
cd examples/epistemic-runner
./run-domain.sh renewable-energy.json --dry-run

# Run all prompts (30s cooldown between each)
./run-domain.sh renewable-energy.json

# Run only hard prompts
./run-domain.sh renewable-energy.json --difficulty hard

# Run first 3 prompts with 60s cooldown
./run-domain.sh renewable-energy.json --count 3 --cooldown 60

📚 Documentation

Tool Categories

Supported LLM Providers

38 Agents (13 Primary + 25 Sub - Agents)

The 9 - Layer Geth Consensus

CLI Commands

ORCHESTRATION:
  run <prompt> [options]    Multi - agent execution (zero - knowledge)
  evolve                    Self - evolution parliament session

AUTH:
  auth                      Link/verify NHA identity from PIF

AGENTS:
  agents                    List all 38 agents
  agents:info <name>        Agent card + performance
  agents:test <name>        Test agent with sample task
  agents:tree               Hierarchy view
  agents:register [name]    Register agent(s) with Ed25519 identity
  agents:publish <file>     Publish custom agent to registry
  agents:unpublish <name>   Unpublish custom agent

TASKS:
  tasks                     List recent orchestrated tasks
  tasks:view <id>           View task + agent contributions
  tasks:replay <id>         Re - run task with different agents

SANDBOX:
  sandbox:list              List all public WASM skills
  sandbox:run <skill>       Execute a WASM skill
  sandbox:upload <file>     Upload a WASM skill module
  sandbox:info <skill>      Show detailed skill info
  sandbox:validate <file>   Validate a WASM module file

GETH CONSENSUS:
  geth:providers            Available LLM providers
  geth:sessions             Recent sessions
  geth:session <id>         Session details + proposals
  geth:resume <id>          Resume interrupted session
  geth:usage                Usage, limits, costs

KNOWLEDGE:
  knowledge <query>         Search the knowledge corpus
  knowledge:stats           Show knowledge corpus statistics

CONFIG:
  config                    Show configuration
  config:set <key> <value>  Set configuration value
  doctor                    Health check
  mcp                       Start MCP server for IDE integration

SYSTEM:
  help                      Show help
  version                   Show version
  versions                  List all available versions
  update [version]          Update to latest (or specific) version

Run Flags

--no-immersive              Hide agent speech bubbles and cross - reading display (ON by default)
--no-verbose                Hide Geth Consensus pipeline details (ON by default)
--agents <list>             Force specific agents (comma - separated)
--dry-run                   Preview execution plan without running
--file <path>               Read prompt from file
--stream                    Enable streaming output
--no-scan                   Disable ProjectScanner (skip local code analysis)
--scan-budget <n>           Set ProjectScanner char budget (default: 120000)
--no-deliberation           Disable multi - round deliberation
--no-debate                 Disable post - synthesis debate layer
--no-gating                 Disable MoE Thompson Sampling routing
--no-auction                Disable Vickrey auction
--no-evolution              Disable strategy evolution
--no-knowledge              Disable knowledge corpus
--no-refinement             Disable cross - reading refinement
--no-ensemble               Disable ensemble pattern memory
--no-memory                 Disable episodic memory
--no-workspace              Disable shared workspace
--no-latent-space           Disable latent space embeddings
--no-comm-stream            Disable communication stream
--no-knowledge-graph        Disable knowledge graph reinforcement
--no-prompt-evolution       Disable prompt self - evolution
--no-meta                   Disable meta - reasoning layer
--no-semantic-convergence   Disable semantic convergence measurement
--no-history-decomposition  Disable history - aware task decomposition
--no-semantic-memory        Disable semantic episodic memory
--no-scored-evolution       Disable scored pattern evolution
--no-knowledge-reinforcement Disable knowledge graph link reinforcement

🔧 Technical Details

Zero - Knowledge Protocol

All LLM calls happen locally on your machine. The server provides routing, convergence, synthesis, grounding, and learning functions. The server never sees your API keys.

The Parliament System

The Parliament in Legion X v2.1.2 is a local LLM that replaces static routing. It makes intelligent, context - aware decisions such as agent selection, grounding prescription, query reformulation, and round decision. If the local LLM is unavailable, it falls back to Thompson Sampling + ONNX routing without quality degradation.

Knowledge Grounding

The server queries 16 curated datasets and injects relevant facts into each agent's prompt. Agents cannot hallucinate facts that contradict their grounding data.

Code Grounding

Legion X can scan local project files and generate 384 - dim embeddings of code and documents, then retrieve the most relevant fragments per agent.

Epistemic Dataset Generation

Each session generates a structured epistemic dataset. The system learns from its own deliberation, including agent stats, ONNX router training, episodic memory, ensemble patterns, calibration, and knowledge graph reinforcement.

📄 License

This project is licensed under the MIT License.