@agishub/mcp 2.1.1 → 2.1.2

package/README.md

@@ -4,6 +4,8 @@
 
 **Pay-per-call tools for AI agents — no signup, no API keys, just USDC.**
 
+### ▶ [Try any tool live in your browser — no install](https://api.agishub.com/try)
+
 Live data & utilities over [x402](https://x402.org) micropayments on **Base**. Your agent
 pays a few tenths of a cent per call from its own wallet; no accounts, no monthly plans.
 
@@ -53,7 +55,7 @@ Discovery:    https://api.agishub.com/openapi.json
 
 ## Tools
 
-**30 tools across 12 services.** Prices are per call in USDC on Base. Tools marked
+**34 tools across 12 services.** Prices are per call in USDC on Base. Tools marked
 **MCP · HTTP** are **free via MCP** and pay‑per‑call via **HTTP x402**; tools marked **HTTP**
 are paid‑only (they touch metered infrastructure, so they run exclusively on the x402 endpoint).
 
@@ -71,11 +73,15 @@ are paid‑only (they touch metered infrastructure, so they run exclusively on t
 | `is_holiday` | Is a date a public holiday in a country (ISO 3166‑1)? Authoritative public‑holiday data. | MCP · HTTP | $0.001 |
 | `list_timezones` | List or search valid IANA timezone names. | MCP | free |
 
-### 🕸️ Web — read any web page as clean markdown
+### 🕸️ Web — read, scrape & extract from any page
 
 | Tool | What it does | Channels | Cost (x402) |
 |------|--------------|:--------:|:-----------:|
 | `extract` | Fetch any public URL and return its main content as clean, token‑efficient markdown (title, headings, links, lists). Optional `render:true` runs a headless browser for JS‑heavy pages / SPAs. Built for RAG. | MCP · HTTP | **$0.004** |
+| `scrape` | Extract specific elements from a JS‑rendered page by CSS selector — text and attributes of every match. Backed by a headless browser, so it works on SPAs. | HTTP | $0.004 |
+| `structured` | AI‑powered structured extraction: give a URL plus a prompt and/or a JSON Schema and get back clean structured JSON (e.g. product name, price, rating). | HTTP | $0.006 |
+| `snapshot` | Capture several representations in one call — rendered HTML, a PNG screenshot, and optional markdown & accessibility tree. | HTTP | $0.008 |
+| `links` | Return every hyperlink on a JS‑rendered page as absolute URLs, with visible‑only and same‑site filters. | HTTP | $0.002 |
 
 ### 🤖 AI — NLP & generation (no external API key)
 

package/dist/stdio.js

@@ -36083,6 +36083,27 @@ var extract = external_exports.object({
   include_images: external_exports.boolean().optional().describe("Keep images as markdown (default false)."),
   max_chars: external_exports.number().int().positive().optional().describe("Truncate the markdown to at most this many characters (sets truncated:true).")
 });
+var scrape = external_exports.object({
+  url: external_exports.string().url().describe("Full http/https URL of the page to scrape."),
+  selectors: external_exports.array(external_exports.string().min(1)).min(1).max(20).describe("CSS selectors to extract, e.g. ['h1', 'a.product', '.price']. Returns the text and attributes of every match per selector.")
+});
+var links = external_exports.object({
+  url: external_exports.string().url().describe("Full http/https URL of the page to read links from."),
+  visible_only: external_exports.boolean().optional().describe("Return only links visible in the rendered layout (default false)."),
+  exclude_external: external_exports.boolean().optional().describe("Drop links pointing to other domains, keeping only same-site links (default false).")
+});
+var structured = external_exports.object({
+  url: external_exports.string().url().describe("Full http/https URL of the page to extract data from."),
+  prompt: external_exports.string().optional().describe("Natural-language instruction of what to extract, e.g. 'the product name, price and rating'. Provide this and/or a schema."),
+  schema: external_exports.record(external_exports.any()).optional().describe("Optional JSON Schema object describing the exact shape of the data to return. When given, the output is constrained to it.")
+});
+var snapshot = external_exports.object({
+  url: external_exports.string().url().describe("Full http/https URL to capture."),
+  formats: external_exports.array(external_exports.enum(["html", "screenshot", "markdown", "accessibilityTree"])).optional().describe("Which representations to return (default ['html','screenshot']). Add 'markdown' and/or 'accessibilityTree' as needed."),
+  full_page: external_exports.boolean().optional().describe("Capture the full scrollable page in the screenshot instead of just the viewport (default false)."),
+  width: external_exports.number().int().positive().optional().describe("Viewport width in pixels (default 1280)."),
+  height: external_exports.number().int().positive().optional().describe("Viewport height in pixels (default 800).")
+});
 
 // src/services/web/core/extract.ts
 var import_node_html_parser = __toESM(require_dist2(), 1);
@@ -36416,6 +36437,117 @@ async function extract2(opts, env2) {
   };
 }
 
+// src/services/web/core/quickactions.ts
+var TIMEOUT_MS = 3e4;
+var QuickActionError = class extends Error {
+};
+function assertPublicHttpUrl2(raw) {
+  let u;
+  try {
+    u = new URL(raw);
+  } catch {
+    throw new QuickActionError("URL inv\xE1lida.");
+  }
+  if (u.protocol !== "http:" && u.protocol !== "https:") {
+    throw new QuickActionError("Solo se admiten URLs http/https.");
+  }
+  const host = u.hostname.toLowerCase().replace(/^\[|\]$/g, "");
+  const isPrivate = host === "localhost" || host.endsWith(".localhost") || host.endsWith(".internal") || host === "::1" || /^(0\.|127\.|10\.|192\.168\.|169\.254\.)/.test(host) || /^172\.(1[6-9]|2\d|3[01])\./.test(host);
+  if (isPrivate) throw new QuickActionError("Host no permitido (red interna).");
+  return u.toString();
+}
+async function callJson(env2, endpoint, body) {
+  const token = env2?.CF_API_TOKEN;
+  const acct = env2?.CF_ACCOUNT_ID;
+  if (!token || !acct) {
+    throw new QuickActionError(
+      "Browser Rendering is not configured (missing CF_API_TOKEN / CF_ACCOUNT_ID)."
+    );
+  }
+  const ctrl = new AbortController();
+  const timer2 = setTimeout(() => ctrl.abort(), TIMEOUT_MS);
+  try {
+    const res = await fetch(
+      `https://api.cloudflare.com/client/v4/accounts/${acct}/browser-rendering/${endpoint}`,
+      {
+        method: "POST",
+        signal: ctrl.signal,
+        headers: { authorization: `Bearer ${token}`, "content-type": "application/json" },
+        body: JSON.stringify(body)
+      }
+    );
+    const data = await res.json().catch(() => null);
+    if (!res.ok || !data || data.success === false) {
+      const detail = data?.errors ? JSON.stringify(data.errors).slice(0, 240) : `HTTP ${res.status}`;
+      throw new QuickActionError(`Browser Rendering ${endpoint} failed: ${detail}`);
+    }
+    return data.result;
+  } catch (e) {
+    if (e instanceof QuickActionError) throw e;
+    throw new QuickActionError(
+      `Browser Rendering ${endpoint} error: ${e instanceof Error ? e.message : String(e)}`
+    );
+  } finally {
+    clearTimeout(timer2);
+  }
+}
+async function scrape2(o, env2) {
+  const url = assertPublicHttpUrl2(o.url);
+  const groups = await callJson(env2, "scrape", {
+    url,
+    elements: o.selectors.map((selector) => ({ selector }))
+  });
+  const elements = (groups || []).map((g) => ({
+    selector: g.selector,
+    count: g.results?.length ?? 0,
+    matches: (g.results || []).map((m) => ({
+      text: (m.text || "").trim(),
+      attributes: Object.fromEntries((m.attributes || []).map((a) => [a.name, a.value]))
+    }))
+  }));
+  return { url, elements, scraped_at: (/* @__PURE__ */ new Date()).toISOString() };
+}
+async function links2(o, env2) {
+  const url = assertPublicHttpUrl2(o.url);
+  const result = await callJson(env2, "links", {
+    url,
+    visibleLinksOnly: !!o.visible_only,
+    excludeExternalLinks: !!o.exclude_external
+  });
+  const list = Array.isArray(result) ? result : [];
+  return { url, count: list.length, links: list, fetched_at: (/* @__PURE__ */ new Date()).toISOString() };
+}
+async function structured2(o, env2) {
+  const url = assertPublicHttpUrl2(o.url);
+  if (!o.prompt && !o.schema) {
+    throw new QuickActionError("Provide a 'prompt' and/or a JSON 'schema' describing what to extract.");
+  }
+  const body = { url };
+  if (o.prompt) body.prompt = o.prompt;
+  if (o.schema) body.response_format = { type: "json_schema", json_schema: o.schema };
+  const data = await callJson(env2, "json", body);
+  return { url, data, extracted_at: (/* @__PURE__ */ new Date()).toISOString() };
+}
+async function snapshot2(o, env2) {
+  const url = assertPublicHttpUrl2(o.url);
+  const formats = o.formats && o.formats.length ? o.formats : ["html", "screenshot"];
+  const body = {
+    url,
+    formats,
+    viewport: { width: o.width || 1280, height: o.height || 800 },
+    screenshotOptions: { fullPage: !!o.full_page, type: "png" }
+  };
+  const r = await callJson(env2, "snapshot", body);
+  const out = { url, formats, captured_at: (/* @__PURE__ */ new Date()).toISOString() };
+  if (r.content != null) out.html = r.content;
+  if (r.markdown != null) out.markdown = r.markdown;
+  if (r.accessibilityTree != null) out.accessibility_tree = r.accessibilityTree;
+  if (r.screenshot) {
+    out.screenshot = { mime: "image/png", base64: r.screenshot, data_uri: `data:image/png;base64,${r.screenshot}` };
+  }
+  return out;
+}
+
 // src/services/web/handlers.ts
 async function extract3(ctx) {
   const { url, render, include_links, include_images, max_chars } = ctx.input;
@@ -36432,10 +36564,26 @@ async function extract3(ctx) {
   }
   return result;
 }
+function scrape3(ctx) {
+  return scrape2(ctx.input, ctx.env);
+}
+function links3(ctx) {
+  return links2(ctx.input, ctx.env);
+}
+function structured3(ctx) {
+  return structured2(ctx.input, ctx.env);
+}
+function snapshot3(ctx) {
+  return snapshot2(ctx.input, ctx.env);
+}
 
 // src/services/web/operations.ts
 var operations2 = {
-  extract: defineOperation(extract, extract3)
+  extract: defineOperation(extract, extract3),
+  scrape: defineOperation(scrape, scrape3),
+  links: defineOperation(links, links3),
+  structured: defineOperation(structured, structured3),
+  snapshot: defineOperation(snapshot, snapshot3)
 };
 
 // src/services/web/index.ts
@@ -36460,7 +36608,7 @@ var screenshot = external_exports.object({
 });
 
 // src/services/render/core/browser.ts
-var TIMEOUT_MS = 3e4;
+var TIMEOUT_MS2 = 3e4;
 var RenderError = class extends Error {
 };
 async function callBinary(env2, endpoint, body) {
@@ -36470,7 +36618,7 @@ async function callBinary(env2, endpoint, body) {
     throw new RenderError("Browser Rendering is not configured (missing CF_API_TOKEN / CF_ACCOUNT_ID).");
   }
   const ctrl = new AbortController();
-  const timer2 = setTimeout(() => ctrl.abort(), TIMEOUT_MS);
+  const timer2 = setTimeout(() => ctrl.abort(), TIMEOUT_MS2);
   try {
     const res = await fetch(
       `https://api.cloudflare.com/client/v4/accounts/${acct}/browser-rendering/${endpoint}`,
@@ -57687,6 +57835,38 @@ var catalog = {
       httpPath: "web-scraper",
       tags: ["web", "scrape", "markdown", "rag", "reader"],
       description: "Fetch any public web page and return its main content as clean, token-efficient Markdown (title, description, headings, links, lists). Set render:true to execute JavaScript first for single-page apps or JS-heavy pages that would otherwise come back empty. Built for RAG and for agents that need to read the contents of a URL."
+    },
+    scrape: {
+      channels: ["http"],
+      pricing: { x402: "$0.004" },
+      visibility: "public",
+      httpPath: "scrape",
+      tags: ["web", "scrape", "selectors", "extract", "html"],
+      description: "Extract specific elements from a JavaScript-rendered page by CSS selector. Give a list of selectors (e.g. 'h1', '.price', 'a.product') and get back the text and attributes of every match. Backed by a headless browser, so it works on SPAs and JS-heavy pages."
+    },
+    links: {
+      channels: ["http"],
+      pricing: { x402: "$0.002" },
+      visibility: "public",
+      httpPath: "links",
+      tags: ["web", "links", "crawl", "urls"],
+      description: "Return every hyperlink on a JavaScript-rendered page as a list of absolute URLs, with options to keep only visible links or only same-site links. Backed by a headless browser. Use it to map a site or seed a crawler."
+    },
+    structured: {
+      channels: ["http"],
+      pricing: { x402: "$0.006" },
+      visibility: "public",
+      httpPath: "extract-json",
+      tags: ["web", "ai", "extract", "structured", "json"],
+      description: "AI-powered structured extraction: give a URL plus a natural-language prompt and/or a JSON Schema, and get back clean structured JSON (e.g. product name, price, rating). Renders the page in a headless browser first, so it works on SPAs."
+    },
+    snapshot: {
+      channels: ["http"],
+      pricing: { x402: "$0.008" },
+      visibility: "public",
+      httpPath: "snapshot",
+      tags: ["web", "snapshot", "html", "screenshot", "markdown"],
+      description: "Capture several representations of a page in one call \u2014 rendered HTML plus a PNG screenshot by default, and optionally Markdown and the accessibility tree. Backed by a headless browser. Saves round-trips when an agent needs both the content and a visual of a page."
     }
   },
   render: {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@agishub/mcp",
-  "version": "2.1.1",
+  "version": "2.1.2",
   "mcpName": "com.agishub/mcp",
   "description": "AgisHub MCP — pay-per-call tools for AI agents (x402, USDC on Base). Timezone, world clock, date math & scheduling; free via MCP, paid via HTTP. No API key.",
   "keywords": [