Grade your AI agents' outputs objectively
AgentGrade scores AI agent outputs for quality, faithfulness, and task success, so you catch regressions before users do. Plug in trajectories, get comparable scores.
AI-powered micro-SaaS — part of the OPC product factory. Web-first (Next.js 14), deployable to Vercel, subscription-ready.
npm install
cp .env.example .env.local # optional: add OPENAI_API_KEY for real AI
npm run dev # http://localhost:3000Without an API key the app runs in Mock mode (returns a demo output).
npm run build
# Vercel: import repo, set framework = Next.js, root = 12_Micro_SaaS出海/agentevalWire Stripe or Waffo in pages/api/checkout (template not included — add per product).
Web-first checkout keeps fees at 2–5% and avoids the 30% app-store cut.
All product-specific text lives in lib/product.ts (name, inputs, system prompt,
pricing, mock). To clone a new product, copy this folder and edit lib/product.ts.
A directory of 100+ vertical AI micro-SaaS tools I built: lxsaihub.com — includes AIActRadar (EU AI Act risk mapping) and AgentRedTeam (AI agent red-teaming).