OpenSpec is an AI-native accessibility engine that detects WCAG violations using axe-core and auto-generates production-ready code fixes, bringing self-healing accessibility to the open web
The Problem: 96% of top websites fail basic accessibility tests, locking out millions of users with disabilities. Existing tools like Lighthouse and axe DevTools only report errors, they never fix them, leaving developers buried under cryptic rule IDs and hours of manual remediation.
The Solution: OpenSpec is a self-healing accessibility engine. It uses axe-core (trusted by Google, Microsoft, and Deloitte) to audit any HTML interface, then leverages Llama 3 via Groq to automatically generate corrected, WCAG 2.1 AA-compliant HTML and ARIA attributes, rendered instantly as a color-coded Before/After diff in your terminal.
Key Features:
Automated Remediation: Moves beyond reporting to deliver side-by-side code diffs for every violation, no manual fixing required.
AI-Generated Fixes: Context-aware corrections for missing alt text, unlabelled inputs, broken ARIA roles, and more.
FOSS-Centric Architecture: LLM-agnostic by design, works with Groq's hosted Llama 3 or any local model. No telemetry, no SaaS, no lock-in.
Zero Infrastructure: Four dependencies, no browser, no Docker, no GPU. Runs anywhere Node.js 18+ runs.
Technical Impact: OpenSpec eliminates the last mile of accessibility tooling, the gap between knowing what's broken and having working code. Independent developers can now ship inclusive interfaces with zero manual auditing overhead.