REM: Reverse-Engineer Any Company's Entire Marketing Machine From One URL

By Prahlad Menon 4 min read

Most “competitor research” is vibes. You open a rival’s site, click around, maybe scroll their Instagram, and come away with a hunch. REM — reverse-engineer-marketing — replaces the hunch with evidence. You paste one URL and it reconstructs the marketing machine behind it: paid ads, organic social, SEO strategy, tracking stack, founders, and funnel — every claim cited to a public source.

It’s MIT-licensed, runs locally, and the baseline mode needs no API keys at all.

What one command gives you

npx reverse-engineer-marketing linear.app

With zero setup and no keys, REM pulls:

  • Tracking stack — Meta, Google, TikTok, LinkedIn, Reddit, X pixels with their IDs, including the ones hidden inside Google Tag Manager containers. It also flags attribution SDKs and visitor-ID tools (RB2B, Warmly, Clearbit) — the presence of those quietly tells you a company is running an outbound sales motion.
  • SEO footprint — sitemap size, programmatic page families (/integrations/{app}, /{a}-vs-{b}, /templates/…), comparison and “alternatives” pages, content cadence, and locales.
  • AI-search stance — whether they publish llms.txt and which AI crawlers they block or allow.
  • Positioning history — their homepage title, H1, and CTAs over time from the Wayback Machine, so you can see every pivot and repositioning.
  • Funnel & monetization — pricing model, free plan/trial, per-seat vs. usage, sales-led signals, affiliate program and commission, social proof.
  • Founders — names from Wikidata, JSON-LD, and the about page, ranked by confidence.
  • App & community signals — App Store ratings and update cadence, Hacker News launches, Product Hunt.

Add an ANTHROPIC_API_KEY and it writes the whole thing up as a strategy report. Add a Scrape Creators key and the paid-ads modules light up:

  • Meta — every ad including paused ones, and critically, the longest-running active ads — the proven winners a competitor keeps paying to run — with their hooks, formats, CTAs, and landing pages.
  • Google — who else is bidding on the brand name (affiliates and resellers piggybacking on their traffic).
  • LinkedIn & Apple Search Ads, plus copycat apps advertising under the same name.

Track changes over time

Run it again next week and diff:

rem diff calai.app

You get the deltas — new ads launched, follower growth per platform, sitemap expansion, channels that switched on or off, and homepage repositioning. It turns competitive research from a one-time snapshot into a monitoring feed.

The part that makes it genuinely useful: it’s an MCP server

REM isn’t just a CLI — it’s a Model Context Protocol server. That means Claude Code, Codex, Cursor, or any MCP client can run it directly, and with MCP you don’t even need an Anthropic key: your agent writes the report itself using the plan you’re already paying for.

npx reverse-engineer-marketing setup   # detects Claude Code / Codex, shows the plan, asks first

The security design is thoughtful. When REM shells out to a subscription engine, it strips ANTHROPIC_API_KEY and OPENAI_API_KEY from that process’s environment so your logged-in plan is used rather than a billable key, runs it in a scratch folder with your other MCP servers and hooks disabled, and binds the local web server to 127.0.0.1 only. It reads public pages and public ad-transparency libraries — nothing gated or private.

Why this matters

The ad transparency libraries (Meta, Google, Apple’s EU Ad Repository) are a goldmine that almost nobody mines systematically, because doing it by hand across a dozen surfaces is tedious. REM automates the tedium and — this is the important part — cites every finding to a URL you can open yourself. No hallucinated “they’re probably spending six figures on TikTok.” Just: here’s the pixel, here’s the ad that’s been running 180 days, here’s the comparison page they built to rank against you.

If you do any competitive positioning, growth, or founder-led marketing, it’s worth the five minutes to run it on your top three competitors — and then on yourself, to see what your own stack reveals to anyone who looks.

npx reverse-engineer-marketing demo   # a recorded teardown, no keys needed

Repo: github.com/dansugc/reverse-engineer-marketing