The Vision

Transform HexFrame from a knowledge mapping tool into an AI orchestration platform where the hexagonal formalism directly drives multi-agent workflows, context management, and system creation.

Core Innovation: HexFrame as AI Operating System

The Jay Agent Architecture

Jay = The system creation agent that helps users build and improve systems using HexFrame formalism:

Center: Help user create a system
├─ 1. Clarify: What system? → [know/ask user]
├─ 2. Foundation: Existing base? → [create/build upon]
├─ 3. Activation: Already using? → [find MVP/continue]
├─ 4. Validation: Serving goal? → [adjust/proceed]
├─ 5. Timing: Improve now? → [yes/reflect]
└─ 6. Evolution: How to improve? → [execute]

Each step has decision gates: does Jay know the answer or ask the user?

HexFrame Formalism for AI Orchestration

1. Plans as Frames

• Goal = Center tile (what we want to achieve/decide)
• Steps = Child tiles (how to achieve the goal)
• Context = Parent/sibling tiles (relevant background)

2. Color-Based Context Engineering

• Same color siblings: Single agent maintains context across tasks
• Tint inheritance: Parent reasoning about child context inclusion
• Lightness gradients: Context filtering/abstraction levels

3. Edge-Based Interaction Specs

• Sibling edges: Data flow, synchronization points
• Parent-child edges: Context inheritance rules, abstraction boundaries

4. Tile Composition for Reusable Context

• Systems can reference shared context tiles
• Version-controlled system knowledge
• Clear dependency graphs for complex orchestration

Implementation Phases

Phase 1: Prompts as Executable Tiles (Immediate Value)

Goal: Make tiles executable with "Run" buttons

• Tiles containing prompts become runnable
• Children provide context for execution
• Immediate value: persistent, reusable, shareable prompts
• Differentiation: Unlike ChatGPT/Claude, prompts are persistent, composable, version-controlled

Phase 2: Jay Integration & Tool Use

Goal: Implement Jay's system creation workflow

• Add CRUD tools for tiles from chat
• Jay creates/updates tiles as it helps users
• Emerging maps ARE the system documentation
• Value: AI that builds structured documentation while helping

Phase 3: Multi-Agent Router

Goal: Route conversations to specialized agents

Center: Hexframe Assistant
├─ Jay: System creation and improvement
├─ Explorer: Navigate existing systems
├─ Teacher: Explain HexFrame concepts
└─ Builder: Technical implementation help

Phase 4: Advanced Orchestration

Goal: Full HexFrame formalism leverage

• Color-based agent routing and context management
• Edge-based workflow orchestration
• Tile composition for complex system interactions
• Multi-iteration agent workflows with persistent plans

Success Criteria

Immediate (Phase 1-2)

• Users can execute prompts directly from tiles
• Jay helps users build systems while creating documentation
• Value proposition clear vs existing AI tools

Medium-term (Phase 3-4)

• Multi-agent conversations feel natural and productive
• Users create complex systems through conversation
• HexFrame becomes the interface layer for AI orchestration

Long-term Vision

• System thinkers become visionaries through AI amplification
• HexFrame formalism proves as the optimal AI coordination protocol
• Users build living systems, not just documentation

The Transformation

From: "Knowledge mapping tool with chat"
To: "AI orchestration platform where spatial relationships drive agent behavior"

The hexagonal map stops being documentation and becomes the computational substrate for AI coordination - the visual representation IS the execution plan.

Dependencies

• Milestone 1: Must complete dogfooding to validate workflow approach
• MCP Integration: Required for tile CRUD from chat
• User Experience: Simple onboarding for non-technical users

This milestone represents the core innovation that makes HexFrame fundamentally different from existing AI tools - turning spatial organization into computational orchestration.