Agent Audit runs Lighthouse against any URL and converts the output into structured fix packs your coding agent can execute. Each fix pack includes the failing audit evidence, CSS selectors to locate the issue in your repo, implementation steps, and acceptance criteria to verify the fix.
Covers performance (Core Web Vitals), accessibility, technical SEO, LLM visibility, and llms.txt readiness. Works with Claude Code, Cursor, GitHub Copilot, and Codex.
Overview
What it does
Agent Audit runs Lighthouse against any URL and converts the output into structured fix packs your coding agent can execute — instead of a raw performance report.
Each fix pack includes:
- The failing audit evidence (CSS selectors, URLs)
- Repo search hints to locate the issue in your codebase
- Step-by-step implementation instructions
- Acceptance criteria to verify the fix by re-running the audit
What it audits
- Performance — FCP, LCP, CLS, TBT (mobile + desktop, 3-run medians)
- Accessibility — WCAG issues with failing element selectors
- Technical SEO — canonical, robots, metadata, JSON-LD, Open Graph
- LLM visibility — AI crawler signals, llms.txt readiness
- Links & images — broken links, missing alt text, image optimization
Setup
claude mcp add agent-audit -- npx -y @fullstackdegen/agent-audit
Server Config
{
"mcpServers": {
"agent-audit": {
"command": "npx",
"args": [
"-y",
"@fullstackdegen/agent-audit"
]
}
}
}