html-meta-mcp 1.0.0 → 1.0.1

package/README.md

@@ -18,3 +18,22 @@ node dist/index.js
 ```
 
 Link preview products charge for this. The page is fetched, bounded, and only the meta tags are returned.
+
+<!-- paywall -->
+## Premium tools
+
+These tools need a license key:
+
+* `http_info`
+* `robots`
+
+Buy a key at https://mcp-marketplace.io/server/io-github-mrfentmen-html-meta-mcp and set it in your MCP client config:
+
+```json
+"env": { "MCP_LICENSE_KEY": "mcp_live_..." }
+```
+
+The key is checked against MCP Marketplace, cached for 24 hours, and keeps
+working offline once one check has succeeded. Every other tool on this server
+stays free.
+<!-- /paywall -->

package/dist/api.js

@@ -69,3 +69,96 @@ export async function meta(args) {
         lines.push(`Published: ${published.slice(0, 10)}`);
     return lines.join("\n");
 }
+// ---- muxC tools:
+function normaliseUrl(raw) {
+    const value = raw.trim();
+    if (!value)
+        throw new HtmlmetaError("Provide a URL");
+    return /^https?:\/\//i.test(value) ? value : `https://${value}`;
+}
+async function fetchText(url, accept, timeout = 15000) {
+    let res;
+    try {
+        res = await fetch(url, {
+            headers: { "User-Agent": UA, Accept: accept },
+            redirect: "follow",
+            signal: AbortSignal.timeout(timeout),
+        });
+    }
+    catch (e) {
+        throw new HtmlmetaError(`Could not fetch ${url}: ${e instanceof Error ? e.message : String(e)}`);
+    }
+    if (!res.ok)
+        throw new HtmlmetaError(`HTTP ${res.status} from ${url}`);
+    return { res, body: await res.text() };
+}
+/**
+ * A HEAD request, so the caller can see what the origin says about the resource
+ * without pulling the body. Some origins answer HEAD poorly, so a 405 or 501
+ * falls back to a ranged GET.
+ */
+export async function httpInfo(args) {
+    const url = normaliseUrl(args.url ?? "");
+    let res;
+    try {
+        res = await fetch(url, { method: "HEAD", headers: { "User-Agent": UA }, redirect: "follow", signal: AbortSignal.timeout(15000) });
+    }
+    catch (e) {
+        throw new HtmlmetaError(`Could not reach ${url}: ${e instanceof Error ? e.message : String(e)}`);
+    }
+    if (res.status === 405 || res.status === 501) {
+        res = await fetch(url, { headers: { "User-Agent": UA, Range: "bytes=0-0", Accept: "*/*" }, redirect: "follow", signal: AbortSignal.timeout(15000) });
+    }
+    if (!res.ok)
+        throw new HtmlmetaError(`HTTP ${res.status} from ${url}`);
+    const h = res.headers;
+    const header = (name) => h.get(name) ?? "not reported";
+    return [
+        `URL: ${res.url || url}`,
+        `Status: ${res.status} ${res.statusText} | method: ${res.status === 206 || res.status === 200 ? "GET/HEAD" : "GET"}`,
+        `Content-Type: ${header("content-type")}`,
+        `Content-Length: ${header("content-length")} | Last-Modified: ${header("last-modified")}`,
+        `Server: ${header("server")} | Cache-Control: ${header("cache-control")}`,
+    ].join("\n");
+}
+/** Fetch /robots.txt and report the crawl rules it declares. */
+export async function robots(args) {
+    const url = normaliseUrl(args.url ?? "");
+    const target = new URL(url);
+    const robotsUrl = `${target.origin}/robots.txt`;
+    const { res, body } = await fetchText(robotsUrl, "text/plain");
+    const text = body.slice(0, 120000);
+    const lines = text.split(/\r?\n/);
+    const relevant = [];
+    const agents = [];
+    const sitemaps = [];
+    let current = null;
+    for (const line of lines) {
+        const trimmed = line.trim();
+        if (!trimmed || trimmed.startsWith("#"))
+            continue;
+        const idx = trimmed.indexOf(":");
+        if (idx < 0)
+            continue;
+        const field = trimmed.slice(0, idx).trim().toLowerCase();
+        const value = trimmed.slice(idx + 1).trim();
+        if (field === "user-agent") {
+            current = value.toLowerCase();
+            agents.push(value);
+        }
+        else if (field === "sitemap") {
+            sitemaps.push(value);
+        }
+        else if (current) {
+            relevant.push(`  (for ${current}) ${field}: ${value}`);
+        }
+    }
+    const named = agents.filter((a) => a.includes("*") || a === "").length;
+    return [
+        `robots.txt: ${robotsUrl} (HTTP ${res.status}, ${body.length} bytes)`,
+        `User-agent groups: ${agents.length}${named ? " (includes *)" : ""}`,
+        `Rules declared: ${relevant.length}`,
+        relevant.length ? relevant.slice(0, 20).join("\n") : "  no rules found",
+        sitemaps.length ? `Sitemaps:\n${sitemaps.slice(0, 8).map((s) => "  " + s).join("\n")}` : "Sitemaps: none declared",
+    ].join("\n");
+}

package/dist/license.js

@@ -0,0 +1,79 @@
+// Premium tools on this server need a license key bought from MCP Marketplace.
+// Buyers put the key in their MCP client config as MCP_LICENSE_KEY. Every tool
+// that is not listed in PREMIUM keeps working without a key.
+const SLUG = "html-meta-mcp";
+// Tools that need a key. Everything else is free.
+const PREMIUM = new Set([
+    "http_info",
+    "robots",
+]);
+const BUY_URL = `https://mcp-marketplace.io/server/io-github-mrfentmen-${SLUG}`;
+// The license check calls the marketplace's verify endpoint directly instead of
+// using @mcp_marketplace/license. That SDK sends no `apikey` header, so every
+// check it makes is rejected by Supabase before it reaches the function. The
+// publishable key below is public by design - it ships in mcp-marketplace.io's
+// own JavaScript and only ever reaches their licence check.
+const DEFAULT_VERIFY_URL = "https://virupvwhtkpkjsiskckg.supabase.co/functions/v1/verify-key";
+const PUBLISHABLE_KEY = "sb_publishable_BuqlW96Ke8C_zzJG-LQv1Q_YHi_r_4h";
+const OK_CACHE_MS = 60 * 60 * 1000;
+const BAD_CACHE_MS = 60 * 1000;
+const seen = new Map();
+const REASONS = {
+    missing_key: "no key is set",
+    invalid_format: "that key is not a valid MCP Marketplace key",
+    not_found: "that key was not found",
+    revoked: "that key has been revoked",
+    rotated: "that key was rotated, use your new one",
+    expired: "that key has expired, renew it",
+    rate_limited: "there were too many checks just now, try again shortly",
+    network_error: "the license server could not be reached",
+};
+function blocked(tool, reason) {
+    const why = REASONS[reason] ?? `the license server answered "${reason}"`;
+    return `"${tool}" needs a license key, but ${why}. ` +
+        `Set MCP_LICENSE_KEY in your MCP client config. Get a key: ${BUY_URL}`;
+}
+async function verify(key) {
+    const res = await fetch(process.env.MCP_LICENSE_VERIFY_URL || DEFAULT_VERIFY_URL, {
+        method: "POST",
+        headers: {
+            apikey: PUBLISHABLE_KEY,
+            Authorization: `Bearer ${PUBLISHABLE_KEY}`,
+            "Content-Type": "application/json",
+        },
+        body: JSON.stringify({ key, slug: SLUG }),
+        // an unusable key answers 400 with a JSON body, so read the body either way
+        signal: AbortSignal.timeout(15000),
+    });
+    const data = (await res.json());
+    if (typeof data?.valid !== "boolean")
+        return { valid: false, reason: "unexpected_response" };
+    return data;
+}
+/**
+ * Returns null when the caller may run the tool, or a short message telling
+ * them how to get a key.
+ *
+ * A good key is remembered for an hour, so a busy session does not hit the
+ * license server on every call and keeps working through a brief outage.
+ */
+export async function premiumRequired(tool) {
+    if (!PREMIUM.has(tool))
+        return null;
+    const key = process.env.MCP_LICENSE_KEY;
+    if (!key)
+        return blocked(tool, "missing_key");
+    const hit = seen.get(key);
+    if (hit && Date.now() - hit.at < (hit.valid ? OK_CACHE_MS : BAD_CACHE_MS)) {
+        return hit.valid ? null : blocked(tool, hit.reason ?? "invalid");
+    }
+    let result;
+    try {
+        result = await verify(key);
+    }
+    catch {
+        result = { valid: false, reason: "network_error" };
+    }
+    seen.set(key, { ...result, at: Date.now() });
+    return result.valid ? null : blocked(tool, result.reason ?? "invalid");
+}

package/dist/server.js

@@ -1,6 +1,7 @@
 import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
 import { z } from "zod";
-import { meta } from "./api.js";
+import { meta, httpInfo, robots } from "./api.js";
+import { premiumRequired } from "./license.js";
 const text = (value) => ({ content: [{ type: "text", text: value }] });
 const textError = (t) => ({ content: [{ type: "text", text: t }], isError: true });
 const READ_ONLY = { readOnlyHint: true, openWorldHint: true };
@@ -20,5 +21,41 @@ export function createServer() {
             return textError(error(e));
         }
     });
+    server.registerTool("http_info", {
+        title: "HTTP info",
+        description: "Status code, content type, length, server and cache headers for a URL, fetched without pulling the body.",
+        inputSchema: z.object({ url: z.string().describe("URL to inspect.") }),
+        annotations: READ_ONLY,
+    }, async (args) => {
+        // ---- paywall: http_info ----
+        const paywallMessage = await premiumRequired("http_info");
+        if (paywallMessage)
+            return { content: [{ type: "text", text: paywallMessage }], isError: true };
+        // ---- paywall: end ----
+        try {
+            return text(await httpInfo(args));
+        }
+        catch (e) {
+            return textError(error(e));
+        }
+    });
+    server.registerTool("robots", {
+        title: "Robots rules",
+        description: "Fetch a site's robots.txt and report the crawl rules and sitemaps it declares.",
+        inputSchema: z.object({ url: z.string().describe("Site URL; /robots.txt is appended to its origin.") }),
+        annotations: READ_ONLY,
+    }, async (args) => {
+        // ---- paywall: robots ----
+        const paywallMessage = await premiumRequired("robots");
+        if (paywallMessage)
+            return { content: [{ type: "text", text: paywallMessage }], isError: true };
+        // ---- paywall: end ----
+        try {
+            return text(await robots(args));
+        }
+        catch (e) {
+            return textError(error(e));
+        }
+    });
     return server;
 }

package/package.json

@@ -1,5 +1,5 @@
 {
-  "version": "1.0.0",
+  "version": "1.0.1",
   "type": "module",
   "repository": {
     "type": "git",