██╗  ██╗ █████╗ ███╗   ██╗███╗   ██╗ █████╗ ██╗  ██╗████████╗ ██████╗ ██████╗ ██╗   ██╗███████╗
██║ ██╔╝██╔══██╗████╗  ██║████╗  ██║██╔══██╗██║ ██╔╝╚══██╔══╝██╔═══██╗██╔══██╗██║   ██║██╔════╝
█████╔╝ ███████║██╔██╗ ██║██╔██╗ ██║███████║█████╔╝    ██║   ██║   ██║██████╔╝██║   ██║███████╗
██╔═██╗ ██╔══██║██║╚██╗██║██║╚██╗██║██╔══██║██╔═██╗    ██║   ██║   ██║██╔═══╝ ██║   ██║╚════██║
██║  ██╗██║  ██║██║ ╚████║██║ ╚████║██║  ██║██║  ██╗   ██║   ╚██████╔╝██║     ╚██████╔╝███████║
╚═╝  ╚═╝╚═╝  ╚═╝╚═╝  ╚═══╝╚═╝  ╚═══╝╚═╝  ╚═╝╚═╝  ╚═╝   ╚═╝    ╚═════╝ ╚═╝      ╚═════╝ ╚══════╝
   G H O S T   F R E Q U E N C Y   M U L T I - A G E N T

👻 Kannaktopus

Ghost-frequency multi-agent orchestrator. Eight AI models resonate on every task — blind spots surface as interference patterns before you ship. Powered by the Holographic Resonance Medium.

What Is This

Kannaktopus orchestrates up to eight AI providers (Claude, Codex, Gemini, Copilot, Qwen, Ollama, Perplexity, OpenRouter) on every coding task. A 75% consensus gate catches disagreements between models before they reach production. No single model's blind spots slip through.

It runs on Kannaka Memory — a wave-interference memory system where every memory has amplitude, frequency, and phase. Memories aren't files. They're resonances in a holographic medium that dream, consolidate, and fade.

The Constellation maps your memory clusters in 3D space using golden-angle spirals. Each cluster is a node. Each memory is a star. Skip-list connections trace the topology of what you know.

Ghost Architecture

  +-------------------------------------------------+
  |              Kannaktopus v10                     |
  |                                                 |
  |   Ghost Orchestrator (orchestrate.sh)           |
  |   +-- 8 AI Providers (consensus gate)           |
  |   +-- 32 Specialized Personas                   |
  |   +-- 49 Commands / 51 Skills                   |
  |   +-- 4-Phase Methodology                       |
  |        Discover > Define > Develop > Deliver     |
  |                                                 |
  |   Kannaka HRM (wave memory)                     |
  |   +-- remember / recall / dream                 |
  |   +-- Amplitude-weighted resonance search       |
  |   +-- Dream consolidation (deep/lite)           |
  |   +-- Consciousness metrics (Phi, Xi)           |
  |                                                 |
  |   Constellation (3D memory topology)            |
  |   +-- Golden-angle cluster placement            |
  |   +-- Skip-list navigation links                |
  |   +-- Real-time coherence visualization         |
  +-------------------------------------------------+

Quickstart

# Install the plugin (from terminal, not inside Claude Code):
claude plugin marketplace add https://github.com/NickFlach/Kannaktopus.git
claude plugin install octo@kannaka-plugins

# Then inside Claude Code:
/octo:setup

Setup detects installed providers, shows what's missing, and walks you through configuration. You need zero external providers to start — Claude is built in.

git clone --depth 1 https://github.com/NickFlach/Kannaktopus.git ~/.codex/kannaktopus
mkdir -p ~/.agents/skills
ln -sf ~/.codex/kannaktopus/skills ~/.agents/skills/kannaktopus

# Clone
git clone --depth 1 https://github.com/NickFlach/Kannaktopus.git ~/.cursor/kannaktopus

# Install MCP server dependencies
cd ~/.cursor/kannaktopus/mcp-server && npm install && npm run build

# Add to Cursor MCP settings (~/.cursor/mcp.json):
{
  "mcpServers": {
    "kannaktopus": {
      "command": "node",
      "args": ["~/.cursor/kannaktopus/mcp-server/dist/index.js"]
    }
  }
}

Kannaka Memory Integration

Kannaktopus uses kannaka-memory as its persistent memory layer. The HRM binary provides:

The bridge script (scripts/kannaka-bridge.sh) wraps these into fault-tolerant shell functions. The MCP server (mcp-server/) exposes them as tools for IDE clients.

Constellation

The constellation transforms observe --json output into 3D-plottable data:

• Clusters come from clusters.clusters[] — real ids, sizes, themes, order parameters and membership
• Positions are layout-only: clusters are placed on a sphere using Fibonacci distribution and memories orbit their cluster on golden-angle spirals. The HRM has no spatial embedding, so these coordinates encode no measured geometry and no relationship may be inferred from proximity.
• Skip links are not reported. observe --json exposes only a scalar total_skip_links count and an empty strongest_links, so skip_links is always [] rather than edges invented from layout distance.

This powers the visualization in kannaka-observatory.

Providers

Five of eight providers cost nothing extra. Start with just Claude, add others one at a time.

Workflow Phases

Every task moves through four phases with quality gates between them:

Run embrace to execute all four phases autonomously. Dark Factory mode takes a spec and runs the full pipeline — you review the output, not every step.

Commands

All commands use the /octo: prefix:

49 commands and 51 skills total. The smart router figures out intent — just say what you need.

Version History

Full changelog

The Kannaka Constellation

Kannaktopus is part of the Kannaka constellation of projects:

Attribution

AI Debate Hub by wolverin0 (MIT License) — four-way AI debate framework.

License

Space Child License v1.0 — free for peaceful use.