AgentScore by Fern

AgentScore evaluates how well company documentation sites work with AI agents, powered by the afdocs npm package.

Architecture

• Next.js 14 App Router — static pages for 30 company scores, API routes for on-demand scoring
• afdocs — fetches and evaluates documentation sites (requires Node.js runtime)
• Vercel — hosting with static generation at build time

Local Development

npm install
npm run dev

Open http://localhost:3000.

Generate Scores

Pre-generate score data for all 30 companies:

npm run generate-scores

This runs afdocs against each company's documentation site and writes results to data/scores/.

Deploy to Vercel

Via CLI

npx vercel --prod

Via Git

Push to main and Vercel will automatically build and deploy.

Project Structure

app/                  # Next.js App Router pages and layouts
  api/                # API routes (Node.js runtime)
  company/[slug]/     # Dynamic company score pages
data/
  scores/             # Pre-generated score JSON files
scripts/
  generate-scores.ts  # Score generation script
public/               # Static assets (logos, images)

Environment Variables

See .env.example. Currently no env vars are required — afdocs runs without API keys.