x-search

name	x-search
description	X.com (Twitter) search & data extraction — cari post dengan advanced operators, extract replies, scroll pagination. Browser-based via Patchright MCP.
triggers	["cari di x","search x","search twitter","x search","twitter search","research x","cari data x","cari tweet","x-research"]
tools	["mcp_patchright_browser_navigate","mcp_patchright_browser_click","mcp_patchright_browser_snapshot","mcp_patchright_browser_run_code","mcp_patchright_browser_wait_for","mcp_patchright_browser_scroll","mcp_patchright_browser_tabs"]

X.com Search & Data Extraction

Search X.com (Twitter), extract posts with full reply threads via Patchright MCP.

Prerequisites

• Patchright daemon running (port 9877)
• Profile logged in to X.com
• Default profile: rickicode

Search URL Patterns

Filter	URL
Top	https://x.com/search?q={query}&src=typed_query
Latest	https://x.com/search?q={query}&src=typed_query&f=live
People	https://x.com/search?q={query}&src=typed_query&f=user
Media	https://x.com/search?q={query}&src=typed_query&f=image

Advanced Search Operators

• "exact phrase" — exact match
• from:username — posts by specific user
• to:username — replies to specific user
• since:2026-01-01 / until:2026-06-30 — date range
• min_faves:1000 / min_retweets:100 / min_replies:50 — engagement threshold
• filter:links / filter:images / filter:videos — content type
• filter:verified — only verified accounts
• -filter:replies — exclude replies
• lang:id / lang:en — language filter

Extract Posts with Card Links (run_code) — VERIFIED WORKING

async (page) => {
  await page.waitForTimeout(5000);
  return await page.evaluate(() => {
    const articles = document.querySelectorAll('article[data-testid="tweet"]');
    return [...articles].slice(0, 10).map(article => {
      const userNameEl = article.querySelector('[data-testid="User-Name"]');
      const links = userNameEl?.querySelectorAll('a') || [];
      const textEl = article.querySelector('[data-testid="tweetText"]');
      const timeEl = article.querySelector('time');
      // Extract card links (link preview cards) — these have actual external URLs
      const cards = article.querySelectorAll('[data-testid="card.wrapper"]');
      const cardLinks = [...cards].map(c => {
        const a = c.querySelector('a');
        return a ? a.href : null;
      }).filter(Boolean);
      return {
        authorHandle: links[1]?.textContent?.trim() || '',
        text: textEl?.textContent?.trim() || '',
        tweetUrl: timeEl?.closest('a')?.href || '',
        cardLinks: cardLinks,
      };
    });
  });
}

Important: Most X posts about "AI free credits" don't have direct URLs in text. The actual external links are in card preview links ([data-testid="card.wrapper"]). These are t.co shortened URLs that redirect to the real destination. Always extract cardLinks for research/promo hunting.

async (page) => {
  await page.waitForTimeout(5000);
  return await page.evaluate(() => {
    const articles = document.querySelectorAll('article[data-testid="tweet"]');
    return [...articles].map(article => {
      const userNameEl = article.querySelector('[data-testid="User-Name"]');
      const links = userNameEl?.querySelectorAll('a') || [];
      const textEl = article.querySelector('[data-testid="tweetText"]');
      const timeEl = article.querySelector('time');
      const replyBtn = article.querySelector('[data-testid="reply"]');
      const retweetBtn = article.querySelector('[data-testid="retweet"]');
      const likeBtn = article.querySelector('[data-testid="like"]');
      const viewsEl = article.querySelector('a[href*="/analytics"]');
      const imgs = article.querySelectorAll('img[src*="pbs.twimg.com/media"]');

      return {
        authorName: links[0]?.textContent?.trim() || '',
        authorHandle: links[1]?.textContent?.trim() || '',
        authorUrl: links[0]?.href || '',
        verified: !!article.querySelector('img[alt="Verified account"]'),
        text: textEl?.textContent?.trim() || '',
        timestamp: timeEl?.getAttribute('datetime') || '',
        timeDisplay: timeEl?.textContent?.trim() || '',
        url: timeEl?.closest('a')?.href || '',
        replies: replyBtn?.getAttribute('aria-label') || '',
        retweets: retweetBtn?.getAttribute('aria-label') || '',
        likes: likeBtn?.getAttribute('aria-label') || '',
        views: viewsEl?.textContent?.trim() || '',
        images: [...imgs].map(i => i.src),
        hasVideo: !!article.querySelector('video, [data-testid="videoPlayer"]'),
        hasQuote: !!article.querySelector('[data-testid="quoteTweet"]'),
      };
    });
  });
}

Extract Replies from Post Detail — VERIFIED WORKING

Navigate to post URL, then extract main post + replies:

async (page) => {
  // Navigate to post first, then:
  await page.waitForTimeout(5000);
  // Scroll to load more replies
  for (let i = 0; i < 3; i++) {
    await page.evaluate(() => window.scrollBy(0, 1000));
    await page.waitForTimeout(2000);
  }
  return await page.evaluate(() => {
    const articles = document.querySelectorAll('article[data-testid="tweet"]');
    return [...articles].map((article, idx) => {
      const userNameEl = article.querySelector('[data-testid="User-Name"]');
      const links = userNameEl?.querySelectorAll('a') || [];
      const textEl = article.querySelector('[data-testid="tweetText"]');
      const timeEl = article.querySelector('time');
      const likeBtn = article.querySelector('[data-testid="like"]');
      const viewsEl = article.querySelector('a[href*="/analytics"]');
      return {
        authorName: links[0]?.textContent?.trim() || '',
        authorHandle: links[1]?.textContent?.trim() || '',
        verified: !!article.querySelector('img[alt="Verified account"]'),
        text: textEl?.textContent?.trim() || '',
        timestamp: timeEl?.getAttribute('datetime') || '',
        likes: likeBtn?.getAttribute('aria-label') || '',
        views: viewsEl?.textContent?.trim() || '',
        isMainPost: idx === 0,
      };
    });
  });
}

Scroll Pagination — VERIFIED WORKING

async (page, maxScrolls = 5) => {
  let allTweets = [];
  const seen = new Set();
  for (let i = 0; i < maxScrolls; i++) {
    const tweets = await page.evaluate(() => {
      return [...document.querySelectorAll('article[data-testid="tweet"]')].map(a => ({
        url: a.querySelector('time')?.closest('a')?.href || '',
        text: a.querySelector('[data-testid="tweetText"]')?.textContent?.trim()?.substring(0, 100) || '',
      }));
    });
    for (const t of tweets) {
      if (t.url && !seen.has(t.url)) {
        seen.add(t.url);
        allTweets.push(t);
      }
    }
    await page.evaluate(() => window.scrollBy(0, 2000));
    await page.waitForTimeout(2000);
  }
  return { total: allTweets.length, tweets: allTweets };
}

Pitfalls

1. Rate limiting: X may show "Rate limit exceeded". Wait 3-5s between navigations.
2. Login wall: Search requires login. Profile must be logged in.
3. Ref staleness: After navigation, always get fresh snapshot.
4. Infinite scroll: Use window.scrollBy() to load more results.
5. Verified badge: img[alt="Verified account"].
6. Post URL: /{user}/status/{id}.
7. Engagement labels: "291 Replies. Reply", "4781 reposts. Repost", "27692 Likes. Like".
8. Quote tweets: Nested structure — extract both outer and inner.
9. t.co link resolution: Card links use t.co shortened URLs. Always resolve with requests.head(url, allow_redirects=True) to get the actual destination URL. Filter out t.co/twitter.com/x.com from final candidates.
10. Card link extraction: Use [data-testid="card.wrapper"] selector for link preview cards. These contain the actual external URLs behind t.co shortlinks.
11. Tab loading: After navigate, X takes 3-5s to load. Use waitForTimeout(5000).
12. Tab title mismatch: Snapshot may show wrong page title initially — wait for load.
13. Tab selection: If multiple tabs open, use tabs(action='select') to switch.
14. MCP client from Python: Use mcp.client.streamable_http.streamablehttp_client + ClientSession — NOT raw HTTP POST. See patchright-browser skill for full pattern.
15. Quote tweets: Nested structure — extract both outer and inner.
16. Tab loading: After navigate, X takes 3-5s to load. Use waitForTimeout(5000).
17. Page title mismatch: Snapshot may show wrong page title initially — wait for load.
18. Tab selection: If multiple tabs open, use tabs(action='select') to switch.