OnyxKraken / Local-first desktop AI agent / Public evaluation mirror

OnyxKraken is a local-first desktop AI agent with screen awareness, memory, voice, and an animated face.

OnyxKraken is built by Mark Barnes / MarkVizion as a public proof system for local-first autonomous AI workflows. It is designed to observe screen context, remember business context, route tools, act through desktop workflows, and make agent behavior visible through a character interface.

What OnyxKraken Does

Reads Windows applications through UI Automation and computer vision fallback.
Uses local Ollama models during normal operation instead of cloud API dependence.
Maintains persistent local memory and retrieval for cross-session context.
Supports voice input, text chat, REST endpoints, an animated face, and demo workflows.
Connects to creative production systems including Blender, web face surfaces, and MarkVizion proof pages.

Public Proof

Public evaluation mirror: https://github.com/20TwentyVizion/OnyxKraken-Public
Live face: https://onyxkraken-face.netlify.app/face
Technical authority page: https://markvizion.com/onyxkraken
Computer vision paper: https://markvizion.com/papers/how-i-built-a-local-ai-agent-with-computer-vision

Release Status

The public evaluation mirror is available for inspection. The broad commercial release channel, pricing, onboarding, and support path should be treated as under review until deliberately finalized.