searchfetch 1.0.2 → 2.0.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Max
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -1,81 +1,96 @@
-# SearchFetch
+# SearchFetch (MCP Server)
 
-A fault-tolerant, stealth-enabled Model Context Protocol (MCP) server for web searching and content fetching. Built specifically for AI Agents, it bypasses Google's GDPR consent screens, Cloudflare Turnstile, and converts heavy HTML into clean, token-optimized Markdown.
+A maximum fault-tolerant, stealth-enabled Model Context Protocol (MCP) server for web searching and content fetching. Built specifically for AI Agents (Cursor, Claude Desktop, OpenHands), it completely bypasses bot detection (Cloudflare Turnstile, Datadome), dynamically handles SPAs/React, and converts bloat into token-optimized Markdown.
 
 ## Features
-* **Multi-Engine Search:** Natively supports DuckDuckGo and Google parsing out of the box. DuckDuckGo is set as the preferred default.
-* **Aggressive Base64 / Image Scrubber:** Implements "nuclear" DOM scrubbing prior to parsing. Guaranteed to NEVER pollute your LLM's context window with giant base64 image strings (`data:image/...`).
-* **Stealth CloakBrowser:** Avoids FingerprintJS, reCAPTCHA, and Cloudflare using Chromium C++ patches and humanized mouse movements natively. 
-* **SPA & React Support:** Waits for network idle to ensure modern Single Page Applications fully execute JavaScript and render before extracting content.
-* **Fault Tolerant:** Extracts whatever DOM was successfully loaded even if a massive, clunky page times out mid-render.
-* **Pagination Support:** Fetches massive webpages iteratively via `start_index` and `max_length` without blowing out AI context tokens limits.
-
-## Installation
-
-1. Clone or copy the directory.
-   ```bash
-   git clone https://github.com/maxylev/searchfetch
-   ```
-2. Install dependencies:
-   ```bash
-   npm install
-   ```
-3. Make the main script executable:
-   ```bash
-   chmod +x index.js
-   ```
-4. Link it globally to your system:
-   ```bash
-   npm link
-   ```
-
-## Configuration
-
-Configure your AI tool/IDE (Cursor, Claude Desktop, Opencode, etc.) to point to this server.
-
-### Example `config.json` (Opencode, Cursor):
+* **Maximum Fault Tolerance:** Implements auto-healing browser sessions, grace-period timeouts for clunky SPAs, and network-level aborting of tracking scripts and media.
+* **Stealth Engine:** Powered by CloakBrowser C++ patches + `humanize` logic. Antibot systems score it as a normal browser because it mathematically moves and renders exactly like one.
+* **Nuclear Token Scrubber:** Strips Base64 images, SVGs, scripts, and inline styles out of the DOM *before* Markdown conversion, guaranteeing your LLM context window won't blow out.
+* **Dual Execution Paths:** Natively supports zero-install execution via both Python (`uvx`) and Node.js (`npx`).
+
+---
+
+## Usage & Installation
+
+You do not need to install this repository manually. Configure your agent to use the zero-install commands `npx` or `uvx` depending on your environment.
+
+### Claude Desktop Configuration
+Add the following to your config:
+
+**Option A: Using Python (`uvx` - Recommended)**
 ```json
 {
-  "mcp": {
+  "mcpServers": {
     "searchfetch": {
-      "type": "local",
-      "command":["npx", "searchfetch"],
-      "enabled": true
+      "command": "uvx",
+      "args": ["searchfetch"]
     }
   }
 }
 ```
 
-### Example `claude_desktop_config.json`:
+**Option B: Using Node.js (`npx`)**
 ```json
 {
   "mcpServers": {
     "searchfetch": {
       "command": "npx",
-      "args": ["searchfetch"]
+      "args": ["-y", "searchfetch"]
     }
   }
 }
 ```
 
+### Cursor / IDE Configuration
+Add it via the **MCP panel** in Cursor settings:
+* **Type:** `command`
+* **Command:** `uvx searchfetch` (or `npx -y searchfetch`)
+
+---
+
 ## Available Tools
 
 ### 1. `websearch`
-Searches the web via Google or DuckDuckGo and returns structured snippets.
-* **`query`** (string): Your search query.
-* **`engine`** (string): `"google"` or `"duckduckgo"` (default `"duckduckgo"`).
-* **`max_results`** (number): Number of results to return (default `10`).
+Searches the web using DuckDuckGo or Google. Returns a clean list of titles, URLs, and snippets. Excellent for researching general knowledge, news, and finding URLs.
+
+**Parameters:**
+* **`query`** *(string, required)*: The search query string.
+* **`engine`** *(string, optional)*: Search engine to use. Can be `"duckduckgo"` or `"google"`. Default is `"duckduckgo"`.
+* **`max_results`** *(number, optional)*: Maximum number of results to return. Default is `10`.
+* **`region`** *(string, optional)*: Region and language code to localize search results. 
+  * Examples: `"us-en"`, `"uk-en"`, `"de-de"`. 
+  * For DuckDuckGo, it maps directly. 
+  * For Google, it maps to the `gl` (country) and `hl` (language) query parameters automatically.
+  * Default is `"wt-wt"` (global/US English).
+* **`safe_search`** *(string, optional)*: Safe search filtering mode. 
+  * `"-1"` for Moderate.
+  * `"1"` for Strict. 
+  * `"-2"` for Off. 
+  * Default is `"-1"`. 
+  * *Note: Only applies to DuckDuckGo.*
 
 ### 2. `webfetch`
-Visits a URL as a stealthy human, waits for the JS to render, completely scrubs visual assets/inline styles to save tokens, and returns the markdown content.
-* **`url`** (string): Full HTTP/HTTPS link.
-* **`format`** (string): Set to `"markdown"` (default), `"clean_html"`, or `"raw_html"`.
-* **`start_index`** (number): Pagination offset.
-* **`max_length`** (number): Maximum character length to return per call (default `10000`).
-* **`block_media`** (boolean): Speeds up page loads by ignoring images, videos, and fonts entirely at the network layer (default `true`).
-
-## Debugging
-If you want to debug JSON-RPC shapes locally:
+Fetch and extract the main text content from any webpage. Fully executes JavaScript to load React/SPAs and aggressively strips images/media (including base64) to save context tokens.
+
+**Parameters:**
+* **`url`** *(string, required)*: The full URL of the webpage to fetch (must start with http/https).
+* **`format`** *(string, optional)*: Output format. Set to `"markdown"`, `"clean_html"`, or `"raw_html"`. Default is `"markdown"` (highly recommended to save context tokens).
+* **`start_index`** *(number, optional)*: Character offset to start reading from for pagination. Use this if a document is too large to fit in the context window. Default is `0`.
+* **`max_length`** *(number, optional)*: Maximum characters to return per request. Default is `10000`.
+* **`block_media`** *(boolean, optional)*: Block images, videos, and fonts entirely at the network layer to drastically speed up page loads and dodge tracking pixels. Default is `true`.
+
+---
+
+## Architecture & Contributions
+This repository utilizes a flat dual-manifest file structure (`package.json` and `pyproject.toml` in the root). When committing changes, ensure parity between `index.js` and `server.py` logic.
+
+### Local Development
 ```bash
-npm run inspector
+# Node.js Testing
+npm i
+npm run inspector-js
+
+# Python Testing
+pip install -e .
+npm run inspector-py
 ```

package/index.js

@@ -14,37 +14,46 @@ import { launch, ensureBinary } from "cloakbrowser";
 import * as cheerio from "cheerio";
 import TurndownService from "turndown";
 
-const logger = {
-  info: (msg) => console.error(`[INFO] ${msg}`),
-  warn: (msg) => console.error(`[WARN] ${msg}`),
-  error: (msg, err) => console.error(`[ERROR] ${msg}`, err || ""),
-};
-
-// ==========================================
-// BROWSER LIFECYCLE MANAGEMENT
-// ==========================================
 class BrowserManager {
   constructor() {
     this.browser = null;
+    this.launchPromise = null;
   }
 
   async getBrowser() {
-    if (!this.browser) {
-      logger.info("Launching stealth CloakBrowser instance...");
-      this.browser = await launch({
-        headless: true,
-        humanize: true, // Native C++ bot-bypass patches + human behavior
-        args: ["--disable-blink-features=AutomationControlled", "--no-sandbox"],
+    if (this.browser && this.browser.isConnected()) return this.browser;
+    if (this.launchPromise) return this.launchPromise;
+
+    this.launchPromise = launch({
+      headless: true,
+      humanize: true,
+      args: [
+        "--disable-blink-features=AutomationControlled",
+        "--no-sandbox",
+        "--disable-dev-shm-usage",
+      ],
+    })
+      .then((browser) => {
+        this.browser = browser;
+        this.browser.on("disconnected", () => {
+          this.browser = null;
+        });
+        return browser;
+      })
+      .catch((err) => {
+        throw err;
+      })
+      .finally(() => {
+        this.launchPromise = null;
       });
-    }
-    return this.browser;
+
+    return this.launchPromise;
   }
 
   async close() {
     if (this.browser) {
       await this.browser.close();
       this.browser = null;
-      logger.info("Browser instance securely closed.");
     }
   }
 }
@@ -52,26 +61,23 @@ class BrowserManager {
 const browserManager = new BrowserManager();
 
 const cleanup = async () => {
-  logger.info("Received termination signal. Shutting down browser...");
   await browserManager.close();
   process.exit(0);
 };
 process.on("SIGINT", cleanup);
 process.on("SIGTERM", cleanup);
 
-// ==========================================
-// CORE LOGIC: SEARCH & FETCH
-// ==========================================
+function getGoogleRegionParams(region) {
+  if (!region || region === "wt-wt") return "hl=en&gl=us";
+  const parts = region.split("-");
+  if (parts.length === 2) return `gl=${parts[0]}&hl=${parts[1]}`;
+  return `gl=${region}&hl=en`;
+}
 
 async function executeSearch(query, maxResults, region, safeSearch, engine) {
-  logger.info(
-    `Searching ${engine.toUpperCase()} via Stealth Browser for: "${query}"`,
-  );
-
   const browser = await browserManager.getBrowser();
   const context = await browser.newContext();
 
-  // Inject Google Consent cookie to universally bypass GDPR popups blocking the DOM
   await context.addCookies([
     {
       name: "CONSENT",
@@ -84,7 +90,6 @@ async function executeSearch(query, maxResults, region, safeSearch, engine) {
   const page = await context.newPage();
 
   try {
-    // Optimization: Block heavy/unnecessary resources to make searches lightning fast
     await page.route("**/*", (route) => {
       const type = route.request().resourceType();
       if (["image", "media", "font", "stylesheet"].includes(type)) {
@@ -95,66 +100,76 @@ async function executeSearch(query, maxResults, region, safeSearch, engine) {
     });
 
     const results = [];
-    let searchUrl = "";
-
-    if (engine === "google") {
-      searchUrl = `https://www.google.com/search?q=${encodeURIComponent(query)}&hl=en&gl=us`;
-    } else {
-      searchUrl = `https://html.duckduckgo.com/html/?q=${encodeURIComponent(query)}&kl=${encodeURIComponent(region)}&kp=${encodeURIComponent(safeSearch)}`;
-    }
+    const searchUrl =
+      engine === "google"
+        ? `https://www.google.com/search?udm=web&udm=14&q=${encodeURIComponent(query)}&${getGoogleRegionParams(region)}`
+        : `https://html.duckduckgo.com/html/?q=${encodeURIComponent(query)}&kl=${encodeURIComponent(region)}&kp=${encodeURIComponent(safeSearch)}`;
 
     try {
-      // Use networkidle to ensure JavaScript fully renders organic results or follows hidden redirects
       await page.goto(searchUrl, { waitUntil: "networkidle", timeout: 15000 });
     } catch (e) {
-      if (e.name === "TimeoutError") {
-        logger.warn(`Network idle timeout on search. Extracting loaded DOM...`);
-      } else {
-        throw e;
-      }
+      // Allow partial rendering on timeout
     }
 
     const pageContent = await page.content();
     const $ = cheerio.load(pageContent);
 
     if (engine === "google") {
-      // Google's core organic result selector
-      $("div.g").each((i, el) => {
+      $("h3").each((i, el) => {
         if (results.length >= maxResults) return;
 
-        const titleEl = $(el).find("h3").first();
-        const linkEl = $(el).find("a").first();
-        if (!titleEl.length || !linkEl.length) return;
+        const h3 = $(el);
+        let linkEl = h3.closest("a");
+        if (!linkEl.length) linkEl = h3.find("a");
+        if (!linkEl.length) return;
 
-        const title = titleEl.text().trim();
         let link = linkEl.attr("href") || "";
 
-        // Handle Google relative redirect links
+        if (!link || (link.startsWith("/") && !link.startsWith("/url?q=")))
+          return;
+        if (
+          link.includes("google.com/search") ||
+          link.includes("support.google.com")
+        )
+          return;
+
         if (link.startsWith("/url?q=")) {
           try {
             link = decodeURIComponent(link.split("/url?q=")[1].split("&")[0]);
           } catch (e) {}
         }
 
-        // Isolate snippet text safely
-        const cloned = $(el).clone();
-        cloned.find("h3, a, script, style, cite").remove();
-        const snippet = cloned.text().replace(/\s+/g, " ").trim();
+        const title = h3.text().trim();
+        if (!title) return;
+
+        let snippet = "";
+        let parent = h3.parent();
+        while (parent.length && parent.prop("tagName") !== "BODY") {
+          const snippetEl = parent.find(
+            "div.VwiC3b, div[style*='-webkit-line-clamp'], div.yXK7lf, div.Uroaid",
+          );
+          if (snippetEl.length) {
+            snippet = snippetEl.first().text().replace(/\s+/g, " ").trim();
+            break;
+          }
+          parent = parent.parent();
+        }
 
-        if (title && link && link.startsWith("http")) {
-          results.push({ position: results.length + 1, title, link, snippet });
+        if (link.startsWith("http")) {
+          if (!results.some((r) => r.link === link)) {
+            results.push({
+              position: results.length + 1,
+              title,
+              link,
+              snippet,
+            });
+          }
         }
       });
     } else {
-      // DuckDuckGo selector
       $(".result").each((i, el) => {
         if (results.length >= maxResults) return;
-
         const titleEl = $(el).find(".result__title a");
-        const snippetEl = $(el).find(".result__snippet");
-        if (!titleEl.length) return;
-
-        const title = titleEl.text().trim();
         let link = titleEl.attr("href") || "";
 
         if (link.includes("/l/?uddg=")) {
@@ -164,17 +179,21 @@ async function executeSearch(query, maxResults, region, safeSearch, engine) {
           } catch (e) {}
         }
 
-        const snippet = snippetEl.text().replace(/\s+/g, " ").trim();
-        if (title && link && link.startsWith("http")) {
+        const title = titleEl.text().trim();
+        const snippet = $(el)
+          .find(".result__snippet")
+          .text()
+          .replace(/\s+/g, " ")
+          .trim();
+
+        if (title && link.startsWith("http")) {
           results.push({ position: results.length + 1, title, link, snippet });
         }
       });
     }
 
     if (results.length === 0) {
-      const pageText = $("body").text().replace(/\s+/g, " ").substring(0, 300);
-      logger.warn(`No results found. DOM Sample: ${pageText}`);
-      return `No results found on ${engine}. The search engine might be showing a captcha/consent screen, or the query returned nothing. Try rephrasing or switching engines.`;
+      return `No results found on ${engine}. The engine may have shown a captcha, or the query returned nothing.`;
     }
 
     return (
@@ -193,8 +212,6 @@ async function executeSearch(query, maxResults, region, safeSearch, engine) {
 }
 
 async function executeFetch(url, format, startIndex, maxLength, blockMedia) {
-  logger.info(`Fetching URL: ${url} | Format: ${format}`);
-
   const browser = await browserManager.getBrowser();
   const context = await browser.newContext();
   const page = await context.newPage();
@@ -214,13 +231,7 @@ async function executeFetch(url, format, startIndex, maxLength, blockMedia) {
     try {
       await page.goto(url, { waitUntil: "networkidle", timeout: 15000 });
     } catch (navError) {
-      if (navError.name === "TimeoutError") {
-        logger.warn(
-          `Network idle timeout on ${url}. Extracting partial DOM...`,
-        );
-      } else {
-        throw navError;
-      }
+      // Allow partial rendering on timeout
     }
 
     const pageContent = await page.content();
@@ -231,30 +242,27 @@ async function executeFetch(url, format, startIndex, maxLength, blockMedia) {
     } else {
       const $ = cheerio.load(pageContent);
 
-      // 🚀 NUCLEAR OPTION FOR BASE64 AND TOKENS 🚀
-      // Physically scrub out all tags that harbor base64 strings or waste tokens
       $(
         "script, style, nav, header, footer, noscript, iframe, svg, aside, .advertisement, img, picture, video, audio, canvas, map, area, dialog",
       ).remove();
-
-      // Remove inline styles from EVERY element to prevent background-image base64 leaks
-      $("*").removeAttr("style");
-
-      // Remove data URIs anywhere else in the document
-      $("*").each((i, el) => {
-        const src = $(el).attr("src");
-        if (src && src.startsWith("data:image")) $(el).removeAttr("src");
-      });
+      $("*")
+        .removeAttr("style")
+        .each((i, el) => {
+          const src = $(el).attr("src");
+          if (src && src.startsWith("data:image")) $(el).removeAttr("src");
+        });
 
       if (format === "clean_html") {
         finalContent = $.html();
-      } else if (format === "markdown") {
+      } else {
         const turndownService = new TurndownService({
           headingStyle: "atx",
           codeBlockStyle: "fenced",
         });
-        finalContent = turndownService.turndown($.html());
-        finalContent = finalContent.replace(/\n{3,}/g, "\n\n").trim();
+        finalContent = turndownService
+          .turndown($.html())
+          .replace(/\n{3,}/g, "\n\n")
+          .trim();
       }
     }
 
@@ -263,14 +271,10 @@ async function executeFetch(url, format, startIndex, maxLength, blockMedia) {
       startIndex,
       startIndex + maxLength,
     );
-    const isTruncated = startIndex + maxLength < totalLength;
-
-    let metadata = `\n\n---\n[Document Info: Showing characters ${startIndex} to ${
-      startIndex + paginatedText.length
-    } of ${totalLength} total.`;
 
-    if (isTruncated) {
-      metadata += ` Use start_index=${startIndex + maxLength} to paginate and read more.`;
+    let metadata = `\n\n---\n[Document Info: Showing characters ${startIndex} to ${startIndex + paginatedText.length} of ${totalLength} total.`;
+    if (startIndex + maxLength < totalLength) {
+      metadata += ` Use start_index=${startIndex + maxLength} to read more.`;
     }
     metadata += `]`;
 
@@ -281,40 +285,41 @@ async function executeFetch(url, format, startIndex, maxLength, blockMedia) {
   }
 }
 
-// ==========================================
-// MCP SERVER INIT & TOOL REGISTRATION
-// ==========================================
+const server = new McpServer({ name: "searchfetch", version: "2.0.0" });
 
-const server = new McpServer({
-  name: "searchfetch",
-  version: "1.3.0",
-});
-
-server.tool(
+server.registerTool(
   "websearch",
-  "Search the web using DuckDuckGo or Google. Returns a clean list of titles, URLs, and snippets. Excellent for researching general knowledge, news, and finding URLs.",
   {
-    query: z.string().describe("The search query string."),
-    engine: z
-      .enum(["duckduckgo", "google"])
-      .default("duckduckgo")
-      .describe("Search engine to use (default: duckduckgo)."),
-    max_results: z
-      .number()
-      .default(10)
-      .describe("Maximum number of results to return (default: 10)."),
-    region: z
-      .string()
-      .default("wt-wt")
-      .describe("Region code (e.g., 'us-en'). Only applies to DuckDuckGo."),
-    safe_search: z
-      .string()
-      .default("-1")
-      .describe(
-        "'-1' for Moderate, '1' for Strict, '-2' for Off. Only applies to DuckDuckGo.",
-      ),
+    title: "Web Search",
+    description:
+      "Search the web using DuckDuckGo or Google. Returns a clean list of titles, URLs, and snippets. Excellent for researching general knowledge, news, and finding URLs.",
+    inputSchema: z.object({
+      query: z.string().describe("The search query string."),
+      engine: z
+        .enum(["duckduckgo", "google"])
+        .default("duckduckgo")
+        .describe(
+          "Search engine to use. Can be 'duckduckgo' or 'google'. Default is 'duckduckgo'.",
+        ),
+      max_results: z
+        .number()
+        .default(10)
+        .describe("Maximum number of results to return. Default is 10."),
+      region: z
+        .string()
+        .default("wt-wt")
+        .describe(
+          "Region and language code to localize search results (e.g., 'us-en', 'uk-en', 'de-de'). For DuckDuckGo it maps directly. For Google, 'us' is country code and 'en' is language. Default is 'wt-wt' (global/US English).",
+        ),
+      safe_search: z
+        .string()
+        .default("-1")
+        .describe(
+          "Safe search filtering mode. '-1' for Moderate, '1' for Strict, '-2' for Off. Default is '-1'. Note: Only applies to DuckDuckGo.",
+        ),
+    }),
   },
-  async ({ query, engine, max_results, region, safe_search }) => {
+  async ({ query, max_results, region, safe_search, engine }) => {
     try {
       const result = await executeSearch(
         query,
@@ -324,46 +329,52 @@ server.tool(
         engine,
       );
       return { content: [{ type: "text", text: result }] };
-    } catch (error) {
-      logger.error("Search Tool failed:", error);
+    } catch (err) {
       return {
-        content: [{ type: "text", text: `Search Error: ${error.message}` }],
+        content: [{ type: "text", text: `Search Error: ${err.message}` }],
         isError: true,
       };
     }
   },
 );
 
-server.tool(
+server.registerTool(
   "webfetch",
-  "Fetch and extract the main text content from any webpage. Fully executes JavaScript to load React/SPAs and aggressively strips images/media (including base64) to save context tokens.",
   {
-    url: z
-      .string()
-      .url()
-      .describe(
-        "The full URL of the webpage to fetch (must start with http/https).",
-      ),
-    format: z
-      .enum(["markdown", "clean_html", "raw_html"])
-      .default("markdown")
-      .describe(
-        "Output format. Markdown is highly recommended to save context tokens.",
-      ),
-    start_index: z
-      .number()
-      .default(0)
-      .describe("Character offset to start reading from for pagination."),
-    max_length: z
-      .number()
-      .default(10000)
-      .describe("Maximum characters to return per request (default: 10000)."),
-    block_media: z
-      .boolean()
-      .default(true)
-      .describe(
-        "Block images/videos/fonts to drastically speed up rendering (default: true).",
-      ),
+    title: "Web Fetch",
+    description:
+      "Fetch and extract the main text content from any webpage. Fully executes JavaScript to load React/SPAs and aggressively strips images/media (including base64) to save context tokens.",
+    inputSchema: z.object({
+      url: z
+        .url()
+        .describe(
+          "The full URL of the webpage to fetch (must start with http/https).",
+        ),
+      format: z
+        .enum(["markdown", "clean_html", "raw_html"])
+        .default("markdown")
+        .describe(
+          "Output format. Set to 'markdown', 'clean_html', or 'raw_html'. Default is 'markdown' (highly recommended to save context tokens).",
+        ),
+      start_index: z
+        .number()
+        .default(0)
+        .describe(
+          "Character offset to start reading from for pagination. Use this if a document is too large to fit in the context window. Default is 0.",
+        ),
+      max_length: z
+        .number()
+        .default(10000)
+        .describe(
+          "Maximum characters to return per request. Default is 10000.",
+        ),
+      block_media: z
+        .boolean()
+        .default(true)
+        .describe(
+          "Block images, videos, and fonts entirely at the network layer to drastically speed up page loads and dodge tracking pixels. Default is true.",
+        ),
+    }),
   },
   async ({ url, format, start_index, max_length, block_media }) => {
     try {
@@ -375,35 +386,22 @@ server.tool(
         block_media,
       );
       return { content: [{ type: "text", text: result }] };
-    } catch (error) {
-      logger.error(`Fetch Tool failed on ${url}:`, error);
+    } catch (err) {
       return {
-        content: [{ type: "text", text: `Fetch Error: ${error.message}` }],
+        content: [{ type: "text", text: `Fetch Error: ${err.message}` }],
         isError: true,
       };
     }
   },
 );
 
-// ==========================================
-// BOOTSTRAP
-// ==========================================
-
 async function main() {
-  logger.info("Initializing MCP Server...");
-
   await ensureBinary();
-
-  // Re-enable STDOUT right before protocol hook-in
   process.stdout.write = originalStdoutWrite;
-
   const transport = new StdioServerTransport();
   await server.connect(transport);
-
-  logger.info("searchfetch successfully connected and listening for requests.");
 }
 
 main().catch((err) => {
-  logger.error("Fatal error during startup:", err);
   process.exit(1);
 });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "searchfetch",
-  "version": "1.0.2",
+  "version": "2.0.0",
   "description": "Fault-tolerant MCP Server for Stealth Web Search and Fetching",
   "type": "module",
   "bin": {
@@ -18,7 +18,8 @@
   },
   "scripts": {
     "start": "node ./index.js",
-    "inspector": "npx @modelcontextprotocol/inspector node ./index.js"
+    "inspector-js": "npx @modelcontextprotocol/inspector node ./index.js",
+    "inspector-py": "npx @modelcontextprotocol/inspector uv run python ./server.py"
   },
   "dependencies": {
     "@modelcontextprotocol/sdk": "^1.29.0",
@@ -27,10 +28,5 @@
     "playwright-core": "^1.59.1",
     "turndown": "^7.2.4",
     "zod": "^4.4.3"
-  },
-  "devDependencies": {
-    "@types/cheerio": "^0.22.35",
-    "@types/node": "^25.6.2",
-    "@types/turndown": "^5.0.6"
   }
 }