Ad Library MCP

Ad Library MCP

MCP server for searching competitor ads and content across Meta Ad Library, Google Ads Transparency Center, and Instagram. Zero auth required — all data sources are public.

Tools

Quick Start

# Clone and install
git clone https://github.com/superpower-growth/ad-library-mcp.git
cd ad-library-mcp
npm install

# Configure
cp .env.example .env
# Edit .env — add your ScrapeCreators API key (required)
# Optionally add Gemini API key for creative analysis

# Run
npm run dev

How resolve-company Works

The most important tool — finds the right brand across platforms using a 2-phase approach that minimizes API credits.

Phase 1 (always runs, ~3 credits):

Phase 2 (only if Phase 1 result is ambiguous, +1-4 credits):

Caching: Results cached for 10 minutes. Second lookup for the same brand is instant (0 credits).

Credit efficiency by brand type:

Results are confidence-ranked (high/medium/low). Active ad count is the tiebreaker — likes can be misleading (a 17K-like unrelated page vs a 3K-like brand with 240 active ads).

Usage Flow

1. resolve-company query="<brand name>"
   → Confidence-ranked matches with Page IDs, Advertiser IDs, active ad counts
   → Also surfaces related pages (influencer/partner pages running ads to the same domain)

2. meta-search-ads page_id="<page_id>" limit=10
   → Newest ads with copy, format, CTA, carousel card count, and Ad Library links

3. meta-get-ad id="<ad_id>" get_transcript=true
   → Full ad detail: carousel cards with per-card content, caption, video transcript, targeting

4. google-search-ads domain="<domain>" get_ad_details=true
   → Headlines, descriptions, impressions, regions, landing pages, Transparency Center links

5. analyze-creatives creatives=[{url: "<image_or_video_url>", type: "video"}]
   → Batch AI analysis: visual descriptions, full verbatim transcript, messaging themes, hooks
   → Images: ~2-3s each (5 concurrent). Videos: ~15-50s each (2 concurrent). Cached per session.

6. ig-search-reels query="<keyword>"
   → Reels ranked by viral ratio (views / followers)
   → BREAKOUT (10x+), STRONG (3-10x), GOOD (1-3x)

7. ig-get-profile handle="<instagram_handle>"
   → Creator profile + top Reels sorted by views with viral ratio

8. ig-get-transcript url="<reel_url>"
   → Full verbatim transcript of a Reel

9. tiktok-search-videos query="<keyword>" or query="#hashtag"
   → Videos ranked by viral ratio (views / followers)

10. tiktok-get-profile handle="<tiktok_handle>"
   → Creator profile + top videos with viral ratio

11. tiktok-get-comments url="<video_url>" keyword="<optional_filter>"
   → Audience comments sorted by likes, filterable by keyword

12. youtube-search-creators query="<niche>" content_type="videos"
   → Creators ranked by median views/subscribers; top videos with engagement % per creator

13. youtube-get-profile handle="<channel_handle>"
   → Channel + top videos with views/likes/comments as % of subscribers

Add to Claude Code

Once deployed, your team can add it with one command:

claude mcp add ad-library --transport http https://ad-library-mcp.fly.dev/mcp

Or add to .mcp.json:

{
  "ad-library": {
    "type": "http",
    "url": "https://ad-library-mcp.fly.dev/mcp"
  }
}

Environment Variables

Deploy to Fly.io

fly launch
fly secrets set SCRAPECREATORS_API_KEY=your-key
fly secrets set GEMINI_API_KEY=your-key  # optional
fly deploy

Data Sources

• Meta Ad Library — via ScrapeCreators API. Ad copy, images, video URLs, CTA, impressions, spend, targeting, carousel cards, transcripts.
• Google Ads Transparency Center — via ScrapeCreators API. Headlines, descriptions, impressions, regions, creative variations, OCR text.
• Instagram — via ScrapeCreators API. Reels search, profile data, user reels with play counts, built-in transcription. Viral ratio ranking (views / followers) for identifying breakout content.
• TikTok — via ScrapeCreators API. Keyword and hashtag video search, creator profiles, video comments with keyword filtering. Viral ratio ranking for identifying breakout content.
• YouTube — via ScrapeCreators API. Niche search across videos and shorts, channel details with subscriber counts, top videos with views/likes/comments, all engagement signals expressed as % of subscribers. Note: YouTube does not expose share count.
• Gemini AI — optional. Batch image/video analysis with visual descriptions, full verbatim transcripts, and messaging theme classification (social-proof, pain-agitate-solve, benefit-driven, etc.).