Ārai
Runtime policy enforcement for AI coding agents via instruction file hooks
Visit arai.taniwha.ai ↗What is Ārai?
Ārai converts instruction files like CLAUDE.md and .cursorrules into enforced guardrails for AI coding assistants. Instead of treating rules as advisory text that models can ignore, it uses native hooks to block prohibited tool calls before they execute and maintains a tamper-evident audit log showing per-rule compliance. The tool works with Claude Code, Grok Build, Codex, and Cursor, plus a git pre-commit gate for any tool.
Designed for developers who want their coding policies actually enforced, Ārai requires no format changes to existing instruction files and runs locally at zero cost. It classifies rules by intent, injects constraints at the point of action, and provides compliance verdicts—whether each rule was obeyed, ignored, or unclear.
Momentum
Daily activity over the last 30 days: upvotes, unique visits, new GitHub stars and Hacker News points.
How do we know it's vibe coded?
vibetrend discovered Ārai automatically. Evidence: GitHub taniwhaai/arai; repo has .claude, AGENTS.md, CLAUDE.md. View source.
Similar vibe-coded apps
#2 this week
#7 this weekOpen-source design workspace that turns coding agents into a local design engine

Manage teams of AI agents to run your business from one dashboard

Free, open-source video editor with auto-captions, effects, and 4K export

Agent development environment for running multiple coding agents in parallel

AI converts documents into natively editable PowerPoint presentations

Open-source video production system directed by AI agents

Open-source AI job search agent that runs locally in your CLI