Facebook Ads Library MCP Server

Facebook Ads Library MCP Server

Skill by ara.so — Marketing Skills collection

This MCP (Model Context Protocol) server enables AI agents to query Facebook's public Ads Library API, retrieve advertising data for brands, and perform AI-powered analysis of ad creative including images and videos. It supports batch processing for efficient multi-brand queries and includes intelligent caching and credit management.

What It Does

• Brand Search: Convert brand names to Meta platform IDs
• Ad Retrieval: Fetch currently running ads for one or multiple brands
• Image Analysis: Analyze ad images for visual elements, text, colors, and composition
• Video Analysis: Deep analysis of video ads using Gemini AI (pacing, storytelling, messaging)
• Batch Processing: Query multiple brands or platform IDs simultaneously with ~88% token savings
• Smart Caching: Reduces API calls and improves performance
• Credit Management: Automatic detection of API credit exhaustion

Installation

Quick Install

git clone https://github.com/proxy-intell/facebook-ads-library-mcp.git
cd facebook-ads-library-mcp

# Run installer
./install.sh  # macOS/Linux
# OR
install.bat   # Windows

Manual Setup

# Clone repository
git clone https://github.com/proxy-intell/facebook-ads-library-mcp.git
cd facebook-ads-library-mcp

# Create virtual environment
python3 -m venv venv
./venv/bin/pip install -r requirements.txt

# Configure environment
cp .env.template .env
# Edit .env and add:
# SCRAPECREATORS_API_KEY=your_key_here
# GEMINI_API_KEY=your_gemini_key_here (optional, for video analysis)

MCP Configuration

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "fb_ad_library": {
      "command": "/full/path/to/facebook-ads-library-mcp/venv/bin/python",
      "args": [
        "/full/path/to/facebook-ads-library-mcp/mcp_server.py"
      ]
    }
  }
}

Cursor (~/.cursor/mcp.json):

{
  "mcpServers": {
    "fb_ad_library": {
      "command": "/full/path/to/facebook-ads-library-mcp/venv/bin/python",
      "args": [
        "/full/path/to/facebook-ads-library-mcp/mcp_server.py"
      ]
    }
  }
}

Replace /full/path/to/ with your actual project path.

API Keys Required

1. ScrapeCreators API (required): Sign up at scrapecreators.com
2. Google Gemini API (optional, for video analysis): Get key at Google AI Studio

Store keys in .env file:

SCRAPECREATORS_API_KEY=your_scrapecreators_key
GEMINI_API_KEY=your_gemini_key

Available MCP Tools

1. get_meta_platform_id

Converts brand name(s) to Meta platform ID(s).

Input: Single brand name (string) or multiple brands (array)

# Single brand
{
  "brand_name": "Nike"
}

# Multiple brands (batch)
{
  "brand_name": ["Nike", "Adidas", "Under Armour"]
}

Output:

{
  "Nike": "123456789",
  "Adidas": "987654321",
  "Under Armour": "456789123"
}

2. get_meta_ads

Retrieves currently running ads for platform ID(s).

Input: Single platform ID (string) or multiple IDs (array)

# Single platform
{
  "platform_id": "123456789"
}

# Multiple platforms (batch)
{
  "platform_id": ["123456789", "987654321"]
}

Output:

{
  "123456789": {
    "ads": [
      {
        "id": "ad_id_123",
        "ad_creative_body": "Ad text content",
        "ad_snapshot_url": "https://...",
        "images": ["https://image1.jpg"],
        "videos": ["https://video1.mp4"]
      }
    ]
  }
}

3. analyze_ad_image

Analyzes visual elements in ad images.

Input:

{
  "image_url": "https://example.com/ad-image.jpg"
}

Output: Detailed analysis of colors, composition, text, people, emotions, and visual elements.

4. analyze_ad_video

Analyzes a single video ad using Gemini AI.

Input:

{
  "video_url": "https://example.com/ad-video.mp4"
}

Output: Comprehensive analysis including pacing, storytelling, brand messaging, visual techniques, and strategic insights.

5. analyze_ad_videos_batch

Analyzes multiple videos in a single API call (~88% token savings).

Input:

{
  "video_urls": [
    "https://example.com/video1.mp4",
    "https://example.com/video2.mp4",
    "https://example.com/video3.mp4"
  ]
}

Output: Array of analyses, one per video, with comparative insights.

6. get_cache_stats

Returns statistics about cached media.

Output:

{
  "total_cached_items": 42,
  "images": 30,
  "videos": 12,
  "total_size_mb": 156.7,
  "oldest_cache": "2025-01-15T10:30:00Z"
}

7. search_cached_media

Searches previously analyzed media.

Input:

{
  "brand": "Nike",          # Optional
  "colors": ["red", "blue"], # Optional
  "has_people": true,       # Optional
  "media_type": "image"     # Optional: "image" or "video"
}

8. cleanup_media_cache

Removes old cached media files.

Input:

{
  "days_old": 30  # Remove cache older than 30 days
}

Common Usage Patterns

Single Brand Analysis

# Agent workflow:
# 1. Get platform ID
platform_id_result = get_meta_platform_id({"brand_name": "Nike"})
platform_id = platform_id_result["Nike"]

# 2. Get ads
ads_result = get_meta_ads({"platform_id": platform_id})
ads = ads_result[platform_id]["ads"]

# 3. Analyze first video ad
video_url = ads[0]["videos"][0]
video_analysis = analyze_ad_video({"video_url": video_url})

Multi-Brand Competitive Analysis

# Agent workflow for batch processing:
brands = ["Nike", "Adidas", "Under Armour", "Puma"]

# 1. Get all platform IDs at once (batch)
platform_ids = get_meta_platform_id({"brand_name": brands})

# 2. Get all ads at once (batch)
all_platform_ids = list(platform_ids.values())
ads_data = get_meta_ads({"platform_id": all_platform_ids})

# 3. Collect video URLs
video_urls = []
for platform_id, data in ads_data.items():
    for ad in data["ads"]:
        if ad.get("videos"):
            video_urls.extend(ad["videos"][:2])  # First 2 videos per brand

# 4. Batch analyze all videos (huge token savings)
video_analyses = analyze_ad_videos_batch({"video_urls": video_urls})

Filtering and Search

# Find all Nike ads with people in red/white colors
cached_results = search_cached_media({
    "brand": "Nike",
    "colors": ["red", "white"],
    "has_people": True,
    "media_type": "image"
})

# Get fresh data and analyze
for result in cached_results:
    print(f"Ad ID: {result['ad_id']}")
    print(f"Analysis: {result['analysis']}")

Code Examples

Example 1: Basic Brand Ad Check

import json
from mcp import get_meta_platform_id, get_meta_ads

# Get Nike's platform ID
platform_data = get_meta_platform_id({"brand_name": "Nike"})
nike_id = platform_data["Nike"]

# Get their current ads
ads = get_meta_ads({"platform_id": nike_id})
nike_ads = ads[nike_id]["ads"]

print(f"Nike is running {len(nike_ads)} ads")
for ad in nike_ads:
    print(f"- {ad['ad_creative_body'][:100]}...")

Example 2: Competitor Video Strategy Comparison

from mcp import get_meta_platform_id, get_meta_ads, analyze_ad_videos_batch

# Define competitors
brands = ["Coca-Cola", "Pepsi", "Dr Pepper"]

# Get platform IDs (batch)
platform_ids = get_meta_platform_id({"brand_name": brands})

# Get all ads (batch)
all_ids = list(platform_ids.values())
all_ads = get_meta_ads({"platform_id": all_ids})

# Collect video URLs
video_map = {}  # Maps video URL to brand
for brand, pid in platform_ids.items():
    for ad in all_ads[pid]["ads"]:
        if ad.get("videos"):
            for video_url in ad["videos"][:1]:  # First video only
                video_map[video_url] = brand

# Batch analyze
video_urls = list(video_map.keys())
analyses = analyze_ad_videos_batch({"video_urls": video_urls})

# Map results back to brands
brand_strategies = {}
for i, video_url in enumerate(video_urls):
    brand = video_map[video_url]
    if brand not in brand_strategies:
        brand_strategies[brand] = []
    brand_strategies[brand].append(analyses[i])

# Print comparison
for brand, strategies in brand_strategies.items():
    print(f"\n{brand} Video Strategy:")
    print(json.dumps(strategies[0], indent=2))

Example 3: Image Analysis with Filtering

from mcp import get_meta_ads, analyze_ad_image

# Get ads for a platform
ads = get_meta_ads({"platform_id": "123456789"})

# Analyze images
for ad in ads["123456789"]["ads"]:
    if ad.get("images"):
        image_url = ad["images"][0]
        analysis = analyze_ad_image({"image_url": image_url})
        
        # Check if red color is dominant
        colors = analysis.get("dominant_colors", [])
        if any("red" in c.lower() for c in colors):
            print(f"Red-dominant ad found: {ad['id']}")
            print(f"Colors: {colors}")
            print(f"Has people: {analysis.get('has_people', False)}")

Environment Variables

All configuration is stored in .env:

# Required for ad retrieval
SCRAPECREATORS_API_KEY=your_api_key_here

# Optional for video analysis
GEMINI_API_KEY=your_gemini_api_key_here

# Optional cache settings (defaults shown)
CACHE_DIR=./cache
MAX_CACHE_SIZE_MB=1000
CACHE_EXPIRY_DAYS=30

Troubleshooting

API Credits Exhausted

Error: "ScrapeCreators API credits exhausted"

Solution: Top up credits at ScrapeCreators Dashboard. The server will automatically resume once credits are available.

Rate Limit Exceeded

Error: Rate limit messages with wait time

Solution:

• Space out large batch requests
• Use batch operations to reduce total API calls
• Wait the specified time before retrying

Video Analysis Not Working

Problem: Video analysis returns errors or empty results

Check:

1. Ensure GEMINI_API_KEY is set in .env
2. Verify Gemini API key is valid at Google AI Studio
3. Check video URL is accessible
4. Ensure video file size is under Gemini's limits

MCP Server Connection Issues

Problem: Agent can't connect to MCP server

Check:

1. Verify virtual environment is activated and dependencies installed
2. Check MCP config file points to correct Python path (venv/bin/python)
3. Ensure .env file exists with API keys
4. Restart Claude Desktop or Cursor after config changes
5. Check logs in Claude Desktop: ~/Library/Logs/Claude/

Import Errors

Error: ModuleNotFoundError or import failures

Solution:

cd facebook-ads-library-mcp
./venv/bin/pip install -r requirements.txt --upgrade

Cache Issues

Problem: Stale or incorrect cached data

Solution:

# Clean old cache (older than 7 days)
cleanup_media_cache({"days_old": 7})

# Or manually remove cache directory
rm -rf ./cache

Performance Tips

1. Use Batch Operations: Always prefer batch API calls when analyzing multiple brands
2. Leverage Cache: Check search_cached_media before making new API requests
3. Video Batch Analysis: Use analyze_ad_videos_batch instead of multiple single calls (88% token savings)
4. Limit Results: When querying ads, process only what you need (e.g., first 5 ads per brand)
5. Monitor Credits: Check ScrapeCreators dashboard regularly to avoid interruptions

Integration Examples

With LangChain

from langchain.tools import Tool
from mcp import get_meta_platform_id, get_meta_ads

tools = [
    Tool(
        name="get_facebook_ads",
        func=lambda brand: get_meta_ads({
            "platform_id": get_meta_platform_id({"brand_name": brand})[brand]
        }),
        description="Get current Facebook ads for a brand"
    )
]

With Custom Python Scripts

#!/usr/bin/env python3
import sys
import json
from pathlib import Path

# Add MCP server to path
sys.path.insert(0, str(Path(__file__).parent / "facebook-ads-library-mcp"))

from mcp_server import get_meta_platform_id, get_meta_ads

def main():
    brand = input("Enter brand name: ")
    platform_id = get_meta_platform_id({"brand_name": brand})[brand]
    ads = get_meta_ads({"platform_id": platform_id})
    print(json.dumps(ads, indent=2))

if __name__ == "__main__":
    main()