# OnyxKraken Flagship System

Canonical URL: https://markvizion.com/onyxkraken

OnyxKraken is the flagship proof system behind MarkVizion: a local-first AI agent that combines screen awareness, voice, memory, tool routing, creative tool control, and product-grade packaging.

## Thesis

OnyxKraken is the center of gravity because it proves the hardest claim: AI should not only answer, it should perceive, remember, act, and improve inside real workflows.

## Pillars

- Autonomous desktop agent: The core loop is observe, think, act, evaluate, and remember. It uses screen context, voice, local models, persistent memory, planning, and tool routing instead of staying trapped in chat.
- Product-grade infrastructure: The system includes licensing, Gumroad integration, onboarding, security, trust ledgers, signed IPC, tests, documentation, and product packaging concerns.
- Creative production surface: Onyx includes an animated face, mini-app surface, visual workflow tools, stage systems, rigged character tools, timeline editing, dialogue tracks, DJ/beat battle systems, and demo playback.
- Ecosystem router: Onyx is designed to drive other creative systems as modules: BlakVision, EVERA, GameKree8r, WorldBuild, JustEdit, Blender, Unity, Unreal, social tools, and knowledge systems.
- Workflow products: The system contains prebuilt workflows such as game-with-assets, music video, book publishing, marketing bundle, concept-to-game, and album-release pipelines.

## System Modules

- Autonomy daemon
- Command router
- Voice stack
- Persistent memory and RAG
- Self-improvement system
- Multi-agent planning
- Animated personality interface
- Stage and animation studio
- Microdrama production pipeline
- Blender / Unity / Unreal toolkits
- MCP server surface
- License and fulfillment logic

## Public Proof

- Public evaluation mirror: https://github.com/20TwentyVizion/OnyxKraken-Public
- Project page: https://markvizion.com/projects/onyxkraken-ai
- Project brief: https://markvizion.com/project-briefs/onyxkraken-ai.md
- Architecture page: https://markvizion.com/architecture
- Local agent paper: https://markvizion.com/papers/how-i-built-a-local-ai-agent-with-computer-vision

## Honest Limits

- OnyxKraken has a public evaluation mirror prepared around the SKIP grant submission path, but the broad commercial release channel, pricing, and onboarding path should be reviewed before calling it generally available.
- The strongest current public claim is architectural proof and grant-review proof: the system exists, exposes a real CLI, documents its local-first architecture, and demonstrates a coherent desktop-agent direction.
- The next authority step is publishing deeper videos, screenshots, benchmarks, and workflow walkthroughs for the modules already present.