How to build a Claude Code agent that scrapes ad libraries, analyzes competitor messaging, and generates branded PDF reports in minutes.
The output: A 10-20 page competitive analysis with ad screenshots, messaging themes, and strategic recommendations. What used to take days now takes ~5 minutes of human review.
Scrapes all active ads from competitor LinkedIn pages with screenshots.
Pulls Facebook/Instagram ads with creative and copy analysis.
Captures search and display ads from Google's transparency center.
Finds competitors via G2, TrustRadius, and category pages automatically.
Extracts themes, value props, CTAs, and positioning across all ads.
Generates a polished PDF with your branding, ready to share.
Input company URL — Agent scrapes the site to understand product, ICP, and positioning
Find competitors — Searches G2, TrustRadius, and category pages for alternatives
Scrape ad libraries — Uses Browser Base to capture LinkedIn, Meta, and Google ads
Analyze messaging — Extracts themes, CTAs, value props, and creative approaches
Generate PDF — Compiles everything into a branded report with WeasyPrint
Terminal-based Claude that runs commands and accesses files
Cloud browser for screenshots and JavaScript page scraping
Converts HTML/CSS to polished, branded PDFs
Optional: host interactive versions of reports
See what the Competitive Intel agent actually produces.
Two real examples: Snowflake (data warehouses) and Glyphic (AI sales tools).
Here are your example reports:
This workshop was created for Emily Kramer's MKT1 community. It covers how 42 Agency built a reusable agent that compresses multi-day competitive research into minutes of human review. The key insight: agents are just text files with instructions — no code required.
42 Agency builds AI-powered demand gen for B2B companies.
Talk to 42 Agency