Most ad-library tooling stops at how many ads are live. This one reads the creative itself - the messaging in an image, the pacing of a video - and caches what it has already looked at, so comparing four brands in one question is a normal request rather than an expensive one.
A server over the Facebook Ads Library. You name a brand, it resolves the platform ID, pulls the ads that page is currently running, and can read the image and video creative back as structured analysis.
- Resolve one or many brand names to their Meta platform IDs in a single call
- Retrieve the ads running for one or several pages at once
- Analyse an ad image for visual elements, on-image text, colours and composition
- Analyse ad video, one at a time or several in one batched call
- Search media you already analysed by brand, colour, people or media type
- Inspect and clear the local media cache when it grows
Self-hosting needs Python 3.12+ and an ads-data API key in SCRAPECREATORS_API_KEY; video analysis additionally needs a Google Gemini key in GEMINI_API_KEY. The install script creates the virtualenv and the .env for you. A hosted version at useproxy.dev skips both keys.
One command plus a key — npx -y @proxy-intell/facebook-ads-library-mcp, then supply credentials