Quickstart Live

Two commands from zero to your first scored, agent-ready read.

1 · Get a key

Create an account and issue a key in the dashboard — the free tier includes 1,000 reads a month, no card required.

2 · Read a page

$ curl -X POST https://agentread.dev/api/v1/read \
  -H "Authorization: Bearer $AGENTREAD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/pricing"}'

3 · Use the result

{
  "url": "https://example.com/pricing",
  "title": "Pricing",
  "markdown": "# Pricing\n\n…",
  "readScore": 82,
  "hallucinationRisk": "low",
  "flags": [],
  "htmlBytes": 812000,
  "markdownBytes": 8100,
  "tokensBefore": 203114,
  "tokensAfter": 1942,
  "latencyMs": 84,
  "cache": "MISS"
}

Feed markdown to your model; branch on hallucinationRisk before you let an agent quote a price. These are the real field names the engine returns — not a simplified example.

Authentication Live

Bearer tokens, issued per account from the dashboard. Keys start with sk-ar-, are sha-256 hashed at rest, and are shown in full exactly once at creation.

Authorization: Bearer sk-ar-…

/api/v1/read and /api/mcp require this header and 401 without it. The free /api/read and /api/scanendpoints (used by this site's own Playground/ReadScan widgets) stay open, IP-rate-limited instead.

MCP server Live

A real remote MCP server (Streamable HTTP) — no local install, no npx package. Add it to any MCP-capable client with your API key as the bearer token:

{
  "mcpServers": {
    "agentread": {
      "url": "https://agentread.dev/api/mcp",
      "headers": { "Authorization": "Bearer sk-ar-…" }
    }
  }
}

Exposed tools

Read API Live

Free scan & playground endpoints Live

Serve middleware Live

Humans get your site. Verified AI crawlers get the Markdown twin. No published npm package yet, so this is the real, copy-pasteable code rather than a fictional install command:

import { NextResponse, type NextRequest } from "next/server";

const AI_CRAWLERS = ["GPTBot", "ChatGPT-User", "ClaudeBot", "PerplexityBot", "CCBot", "Bytespider"];

export async function middleware(request: NextRequest) {
  const ua = request.headers.get("user-agent") ?? "";
  if (!AI_CRAWLERS.some((c) => ua.includes(c))) return NextResponse.next();

  const res = await fetch("https://agentread.dev/api/v1/read", {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.AGENTREAD_API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({ url: request.url }),
  });
  if (!res.ok) return NextResponse.next();
  const { markdown } = await res.json();
  return new Response(markdown, { headers: { "content-type": "text/markdown" } });
}

export const config = { matcher: "/:path*" };

This exact pattern (crawler UA detection → real distill → Markdown response) is what runs on agentread.dev itself, in src/proxy.ts.

How ReadScore is computed

Fully transparent — starts at 100, then deducts for: low payload reduction, high script count (>25 tags), price/CTA text present in raw HTML but absent from extracted text (JS-only rendering), disabled buy/checkout buttons in markup, lazy-loaded content, and a missing /llms.txt. Every deduction ships as a human-readable flag alongside the score — see src/lib/engine/read.ts for the exact logic.

Rate limits

Roadmap

Not built yet — listed here instead of documented as if callable today:

• MCP tools: batch, map_site, extract_data
• Crawl (whole-domain Markdown corpus)
• Watch (change-detection webhooks)
• llms.txt Studio (auto-generate & host llms.txt / llms-full.txt)
• Agent-traffic analytics dashboard
• Pay-per-crawl monetization for publishers
• Billing / Stripe integration
• Act — semantic agent transactions (long-term)