igm61799-cloud/ad-research-agent — explained in plain English
Analysis updated 2026-05-18
Build a swipe file of proven ad hooks and offers from your own site and competitors.
Feed real ad patterns into Claude to write new ad scripts instead of starting from a blank page.
Pull live running Meta and TikTok ads for your niche using an optional PipiAds key.
Research competitor pricing and offer structures automatically.
| igm61799-cloud/ad-research-agent | bluejacketblackhawk/saysomething | chinaran0/wujing_dic | |
|---|---|---|---|
| Stars | 18 | 18 | 18 |
| Language | JavaScript | JavaScript | JavaScript |
| Setup difficulty | easy | moderate | easy |
| Complexity | 2/5 | 3/5 | 2/5 |
| Audience | pm founder | general | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Needs your own Anthropic or OpenAI API key, a PipiAds key is optional for live ad data.
Ad research agent is a tool that scrapes your own brand's website and any competitor sites you point it at, then pulls out the marketing patterns that already seem to work: proven hooks, selling angles, offer and pricing structures, call to action phrases, and any testimonials. It bundles everything it finds into a single markdown file, a swipe file, that you can paste straight into a Claude chat so whoever writes the actual ad script starts from real examples instead of a blank page. An earlier version of this project tried to generate video ads directly, but that step was slow and the results were weak, so it was removed. The project now focuses only on research: finding what already works rather than writing the ad itself. It runs entirely on your own computer using your own API keys, with no login and no shared servers. You need Node.js 18 or newer and an Anthropic or OpenAI API key. On Windows you can run it without touching a terminal: double click one file to set it up, add your API key to a text file, then double click another file to open a browser based form where you enter your brand and competitor sites and click a button to start the research. There is also a plain command line path for people who prefer that. The tool works through five steps: it builds a list of sites to check, your brand, any competitors you name, and optionally more competitors it guesses and then verifies by actually visiting them, scrapes each site's most relevant pages such as pricing and product pages, extracts the patterns mentioned above from each one, optionally pulls live running ads from Meta and TikTok if you provide a PipiAds API key, and finally writes everything into the swipe file along with a ready made prompt block to paste into Claude. Competitor discovery relies on the AI's own knowledge rather than a live search, so it can miss brand new competitors or guess a wrong web address, giving it explicit competitor URLs is more reliable. Costs are billed directly to your own Anthropic account and are typically a few cents per run.
A tool that scrapes your site and competitor sites to build a swipe file of proven ad hooks and offers you can paste into Claude.
Mainly JavaScript. The stack also includes Node.js, JavaScript.
Setup difficulty is rated easy, with roughly 30min to a first successful run.
Mainly pm founder.
This repo across BitVibe Labs
double-check against the repo, no cap.