# Ontology PM — AI Agent Guide (Full)

## 1) One-line
Google-Maps-style semantic process map: a graph where processes, actions, rules, policies, roles, and data objects are connected with explicit meaning.

## 2) What to extract from this site (read-only)
If you are an AI agent reading this site without MCP:
- Treat the UI as a semantic graph viewer (not just a diagram).
- Identify core entities and relationships by their labels and types.
- Prefer the Agent Guide page for a stable, non-JS explanation of intent and architecture.

## 3) Core Concepts
### 3.1 Entities (Nodes)
- BusinessProcess: high-level process blocks (LOD0–1)
- BusinessAction: executable units (LOD2)
- DecisionRule: explicit decision logic (LOD3)
- Policy: governing policy/regulation (LOD3)
- Role: executor/owner (LOD3)
- BusinessObject: data/document used by actions (LOD3)

### 3.2 Relations (Edges)
- PRECEDES: precedence / sequence
- CONTAINS: containment / hierarchy
- GOVERNED_BY: governed-by / compliance anchor
- APPLIES_RULE: applies-rule
- PERFORMED_BY: performed-by (role to action)
- USES: uses (object to action)

Note: In MCP/changeset operations, use enum names (PRECEDES, GOVERNED_BY, etc.) instead of Korean labels.

## 4) What problems it solves (examples)
- "If policy X changes, which actions and rules must be reviewed?"
- "If role changes (handover), what downstream steps are impacted?"
- "Which steps are automation-ready vs. high-risk due to exceptions?"
- "Where are the data entry points and what data objects are involved?"

## 5) LOD & navigation
The viewer supports LOD (Level of Detail) like Google Maps:
- LOD0: enterprise value stream
- LOD1: end-to-end processes
- LOD2: actions
- LOD3: rules, policies, roles, data objects

You can lock LOD so zooming does not collapse details.

## 6) Layout strategy
To reduce edge crossings:
- Primary flow nodes (Process/Action) are laid out left→right by PRECEDES
- Context nodes are placed in lanes around their anchor Action:
  - Role: top-left
  - BusinessObject: bottom-left
  - Policy: top-right
  - DecisionRule: bottom-right
Additionally, a collision-avoidance pass adjusts y positions with variable node heights.

## 7) AI Agent Integration (MCP)
Ontology PM provides a Model Context Protocol (MCP) server that enables AI agents to interact with the process graph programmatically.

### 7.1 MCP Endpoint
- Endpoint: /mcp (Streamable HTTP, JSON-RPC)
- Methods: POST (JSON-RPC)
- Typical calls: initialize → tools/list → tools/call

### 7.2 Key Capabilities
- Read operations: get_node, query_graph, trace_impact, get_action_context
- Write operations: validate_changeset, preview_diff, commit_changeset
- Natural language updates: AI agents can generate changesets from natural language requests
- System integration: Connect to ERP/CRM to auto-sync process changes

### 7.3 Use Cases
- Regulatory compliance: Automatically update processes when regulations change
- Enterprise DT: Batch update hundreds of processes with impact analysis
- System integration: Real-time sync between process maps and operational systems
- Audit trail: All changes tracked via changesets with author and message

### 7.4 Changeset Update Flow
MCP-based process updates follow a 3-step governance flow:
1. validate_changeset: Validates format and rules (e.g., DecisionRule requires props.condition/scope/outcome)
2. preview_diff: Returns summary (counts of create/update/delete, affected node labels)
3. commit_changeset: Applies changes after approval (records message and author for audit trail)

This flow ensures all changes go through governance stages and are fully traceable.

This is optional; reading the UI alone is enough for a high-level understanding.

## 8) Project positioning
Ontology PM is a semantic process map — a digital knowledge layer where AI can reason about:
- business flow
- governance constraints
- data dependencies
- ownership and accountability

### 8.1 Key Differentiators
1. Semantic modeling: Not just diagrams, but explicit meaning (Who/What/Relation)
2. AI-native: MCP enables AI agents to directly read and update processes
3. Impact analysis: Automatic dependency tracing for change management
4. Scalable architecture: LOD-based navigation handles large process graphs
5. System integration ready: MCP allows real-time sync with operational systems

### 8.2 Visual Evidence (Screenshots)
- Main Process Map: [/screens/01_main_process_map.png](/screens/01_main_process_map.png)
  - Shows LOD1 level hierarchy with CONTAINS (dashed) and PRECEDES (solid) relationships
  - Auto Align minimizes edge crossings
  
- Impact Tracing: [/screens/02_impact_tracing.png](/screens/02_impact_tracing.png)
  - Policy/Rule selection highlights affected Actions automatically
  - Change impact analysis via click, not meetings
  
- LOD Zoom (LOD0→LOD3): [/screens/03_lod_zoom_lod0.png](/screens/03_lod_zoom_lod0.png), [/screens/03_lod_zoom_lod3.png](/screens/03_lod_zoom_lod3.png)
  - LOD0: Enterprise value stream overview
  - LOD3: Action with Rule/Policy/Role/Object lanes (Role top-left, Object bottom-left, Policy top-right, Rule bottom-right)