Ark-Context

The Open Protocol for Context-as-Code

🇰🇷 한국어 설명 (Korean Guide)

This document is written in English for global accessibility. 한국어 설명을 보시려면 위 링크를 클릭해주세요.

---

"Don't just build Agents. Build Identities."

Ark-Context is an open-source framework that distills your "Operating System of Thought" into a structured context that AI can understand. Instead of ephemeral chats, build a permanent, evolving knowledge architecture.

---

🚀 For Everyone (General)

Why Ark-Context?

• Identity Protocol: A standard way to implant your unique philosophy into any AI.
• Ownership: Your context lives on your disk, not in a model's cloud.
• Clarity: Transforms vague thoughts into a structured "Second Brain."
• Zero Fatigue: Stop explaining "who you are" in every new chat. The AI already knows.

Quick Start (30 Seconds)

1. Install

git clone https://github.com/dev-whitecrow/ark-context.git
cd ark-context

2. Run Open this folder in your AI Code Editor (Antigravity, Claude Code, etc.).

3. Onboard Type /onboarding in the AI chat. The agent will guide you to build your Foundation.

---

🛠️ For Engineers (Technical)

The Architecture: Context-SOLID

We applied software engineering principles to knowledge management.

1. SRP (Single Responsibility): Each file holds exactly one semantic unit. High token efficiency.
2. OCP (Open-Closed): Foundation is stable; Execution is extensible. Layered architecture.
3. LSP (Liskov Substitution): Semantic checking ensures tactical docs don't violate strategic core values.
4. ISP (Interface Segregation): AI loads only the relevant context (RAG optimization).
5. DIP (Dependency Inversion): Metrics (Details) depend on Philosophy (Abstractions), not the other way around.

The Core Engine: Multi-Agent System

The system is powered by specialized AI agents working in tandem:

1. The Architect (Distill Skill): Normalizes unstructured Inbox data into pure The_Ark knowledge.
2. The Guardian (Monitoring Skill): Enforces logical integrity and detects contradictions.
3. The Scribe (Log Skill): Immutably records decisions and context for temporal awareness.

Future Proofing

• MCP Ready: Designed to integrate with Anthropic's Model Context Protocol.
• SFT Dataset: A high-quality, self-supervised dataset for fine-tuning your Personal LLM.
• Local First: 100% private, ensuring security for future On-Device AI.

---

Maintained by Project Hur (Whitecrow) "We build the protocol for the AI-Native Era."