Charge AI training crawlers for access (@profullstack/x402-gateway) (#234)

Training crawlers (GPTBot, ClaudeBot, CCBot, meta-externalagent, Bytespider,
Applebot-Extended) get 402 Payment Required with an x402 offer, or the sales
page at /crawl, and a paid pass opens the site for a day. People, search
engines and retrieval crawlers pass through untouched. robots.txt is now
generated from the same lists.


Claude-Session: https://claude.ai/code/session_01YafYxayh7Gqe5MWNNQMev2

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>

Author: ralyodio

.env.example

@@ -114,3 +114,9 @@ GITHUB_APP_SLUG=
 # Playwright owner-flow e2e. Email defaults to anthony+riotcoder@profullstack.com.
 PLAYWRIGHT_AUTH_EMAIL=
 PLAYWRIGHT_AUTH_PASSWORD=
+
+# Crawl gateway (@profullstack/x402-gateway): AI training crawlers pay $1/day over
+# x402. A SCOPED CoinPay key (payments:create) and the EVM address that receives
+# the USDC. Unset = crawlers still get 402, nothing sold.
+COINPAY_X402_KEY=
+CRAWL_PAY_TO=

app/robots.ts

@@ -0,0 +1,9 @@
+import { robotsRoute } from "@profullstack/x402-gateway/next";
+import { gateway } from "@/lib/crawl-gateway";
+
+// Generated from the same crawler lists the gateway enforces: training
+// crawlers are refused everywhere but /crawl (where they can buy a pass),
+// retrieval crawlers are named as welcome, everyone else gets the rules below.
+export const GET = robotsRoute(gateway, {
+  disallow: ["/api/"],
+});

app/robots.txt/route.ts

@@ -0,0 +1,27 @@
+import { createGateway } from "@profullstack/x402-gateway";
+import { x402Proxy } from "@profullstack/x402-gateway/next";
+
+/**
+ * Sells crawl access to AI training crawlers (GPTBot, ClaudeBot, CCBot,
+ * meta-externalagent, Bytespider, Applebot-Extended, ...) by the day over
+ * x402, settled by CoinPay in USDC. People, Googlebot and the retrieval
+ * crawlers behind AI search pass through untouched.
+ *
+ * Runs inside the middleware, so nothing here may import Node-only modules.
+ * The env is read through a non-literal key on purpose: Next inlines
+ * `process.env.NAME` at build time, and these are runtime secrets. Without
+ * COINPAY_X402_KEY and CRAWL_PAY_TO the gateway still answers training
+ * crawlers with 402, just with an empty offer.
+ */
+const env = (name: string) => process.env[name];
+
+export const gateway = createGateway({
+  siteUrl: env("SITE_URL") || env("NEXT_PUBLIC_SITE_URL") || "https://crawlproof.com",
+  siteName: "CrawlProof",
+  coinpay: { apiKey: env("COINPAY_X402_KEY") },
+  payTo: env("CRAWL_PAY_TO"),
+  contact: "mailto:support@crawlproof.com",
+});
+
+/** Resolves to a Response for a refused crawler, or undefined to carry on. */
+export const gate = x402Proxy(gateway);

lib/crawl-gateway.ts

@@ -29,6 +29,7 @@
     "@profullstack/referrals": "^0.1.0",
     "@profullstack/stack": "^0.1.3",
     "@profullstack/x402-client": "^0.2.0",
+    "@profullstack/x402-gateway": "^0.1.0",
     "@supabase/ssr": "^0.10.3",
     "@supabase/supabase-js": "^2.105.4",
     "@types/nodemailer": "^8.0.0",

package-lock.json

@@ -1,10 +1,17 @@
+import { gate } from "@/lib/crawl-gateway";
 import { NextResponse, type NextRequest } from "next/server";
 import { createServerClient, type CookieOptions } from "@supabase/ssr";
 import { trackReferralCode } from "@profullstack/stack/referrals";
 
 type Cookie = { name: string; value: string; options?: CookieOptions };
 
 export async function proxy(request: NextRequest) {
+  // Crawl gateway first: AI training crawlers get 402 Payment Required (or the
+  // sales page at /crawl) unless they present a paid pass. People, Googlebot
+  // and retrieval crawlers fall through to everything below.
+  const answer = await gate(request);
+  if (answer) return answer;
+
   // 308 redirect www.crawlproof.com -> crawlproof.com (preserves method + body).
   const host = request.headers.get("host") ?? "";
   if (host.toLowerCase().startsWith("www.")) {