ontoraness

AI Ontology + Harness system for Claude Code — automate project rules and context injection.

Define your project's architecture, domain, and code style rules as structured YAML ontologies.
The harness compiles them into Claude Code context with minimal tokens, injecting the right rules at the right time.

Quick Start

npx ontoraness onboard

Commands

onboard    Scan project → generate ontology draft → install harness
generate   Compile ontology → CLAUDE.md + docs/ + ESLint rules
validate   Validate ontology schema
report     Show compliance rate, violation patterns, doc usage stats
agent      Delegate tasks to sub-agents (GPT, Gemini)
ui         Run the local ontology and harness UI
features   Manage optional Ontoraness features
docs       Manage project documentation catalog/search
graft      Register and sync external managed workspaces

How It Works

Define rules in .ontoraness/ontology/*.yml and the harness automatically:

• Generates CLAUDE.md — core rules summary (~200 tokens, always loaded)
• Generates docs/ — detailed rules injected per-task into AI context
• Generates ESLint rules — Tier3 rules enforced via .eslintrc.ontoraness.mjs
• Registers Claude Code hooks — pre/post-tool validation and injection

Context Tier System

Tier 1  Always loaded (~200 tokens)    → CLAUDE.md (absolute constraints)
Tier 2  Injected on context match      → docs/ (task-specific guides)
Tier 3  Code-enforced (0 tokens)       → ESLint + PreToolUse hook

Documentation

• Getting Started
• Ontology Schema Reference
• Context Tier System
• Sub-Agents Guide
• CLI Reference

Repository

https://github.com/Reone1/ontoraness