Features
Document ingestion from PDF, HWP, GitHub, expert notes
OCR and CLIP-based parsing
Nine-space MetaOntology structure (Subject, Evidence, Claim, Policy, Lever)
Evidence graph construction with nodes, edges, and evidence
Ontology pack creation with manifest
MCP integration for AI workspaces (Claude Desktop, GPT)
CrabSkill workflow with validator and playbook
Agent deployment for Codex and Claude Code
Pack publishing and marketplace
Private workspace with isolated storage (Pro/Expert)
Reverse ingest from connected MCP tools (Expert)
Webapp, desktop client, and MCP runtime panels
Korean and English language support
Architecture diagram with six-stage pipeline
Semantic knowledge graph construction from enterprise data
Deep contextualisation of structured and unstructured data
Expert validation loops with versioned audit trails
LLM evaluation against accuracy and compliance criteria
Guardrails enforcing consistency from expert benchmarks
Full source traceability and explainable reasoning paths
Model-agnostic flexibility (no vendor lock-in)
Ingests documents, patents, regulations, research, and ERP data
330M+ documents securely ingested across 200K+ evaluated answers
Quantified confidence scores at every layer
Integrates with AWS for regulated deployment
Axion™ product: from data chaos to AI-ready intelligence
Neuralith™ product: enterprise knowledge into an AI engine
RSpace™ product: precision intelligence for complex R&D
Strategic partnership with AWS
