last30days: your market research in one command, ranked by real engagement
The free tool that scans up to 15 platforms (Reddit, X, YouTube, TikTok, Polymarket, GitHub) and hands you a sourced brief in 30 seconds. Ranked by real engagement, not SEO. Install, first search, and marketing use cases.
June 26, 2026 · 6 min read
Why last30days
When you want to know what people are actually saying about a topic, Google shows up late. By the time an article ranks, the conversation moved on to Reddit, X and YouTube three weeks ago. So you do the research by hand: you scroll ten Reddit threads, you read the comments, you open three YouTube videos, you check what is trending on TikTok. An hour later you have a rough idea, but no clean sources and nothing ranked.
last30days flips the problem. It is an AI agent-led search engine that queries around fifteen platforms at once, and ranks results by real engagement: upvotes, likes, views, even real-money bets. You run one command, it resolves your topic to the right accounts and communities, it reads the content from the last 30 days, it merges duplicates and it hands you a sourced brief. Open source, MIT license, and the core sources run without a single API key.
This guide shows you what makes it different from Google, the platforms it scans, how to install it and run your first search, then four concrete ways to use it in marketing.
The Claude AI Lab is my Skool community where I share my Claude systems and the more advanced modules. Entry is free.
Join the Lab →Why it is not Google
Google ranks results by SEO relevance: who has the best optimization, the most backlinks, the best-tuned content. last30days ranks by social relevance: what actually made people react.
In practice, a Reddit thread with 1,500 upvotes is a stronger signal than a blog post nobody read. The engine scores every piece on several factors: upvote and like counts, view numbers, freshness (everything comes from the last 30 days), and for Polymarket the real-money bets that beat any poll. It also caps each author at 3 items, so a single loud voice never takes over.
And it does not just paste links. When the same story shows up on Reddit, on X and on YouTube, it groups it into one cluster instead of repeating it three times. At the end, a second judge picks out the funniest and most viral one-liners for a "Best Takes" section.
You stop searching for what is well optimized, and start searching for what actually matters to a community. That is exactly the signal you need for content, market research or competitive intel.
The platforms it scans
Every network is a walled garden, with its own API and its own auth. last30days bridges all of them, and queries them at once.
The four free sources (Reddit, Hacker News, Polymarket, GitHub) already cover a huge range of cases. You add X, YouTube or TikTok only when you want to push further.
Install it and run your first search
In Claude Code, a single line installs the plugin:
/plugin marketplace add mvanhorn/last30days-skill
On another host (Cursor, Codex, Gemini CLI), you use the skills command:
npx skills add mvanhorn/last30days-skill -g
Then you run a search by simply naming your topic:
/last30days AI marketing for agencies
The engine first resolves your topic to the right accounts, subreddits and channels, then queries every platform in parallel. You can also compare two things, or ask for the competitive view:
/last30days Claude vs ChatGPT for marketing
/last30days HubSpot --competitors
The result saves locally, in Markdown by default, or as a self-contained HTML file (dark mode, ready to share) with --emit=html. An eli5 mode rewrites the whole thing in plain language if you need to simplify.
Everything stays on your machine. No tracking, no analytics, your research goes nowhere. The files land in ~/Documents/Last30Days/.
4 marketing use cases
This is where it becomes a real lever for you.
Find content angles. You give it your topic, the engine surfaces the questions, frustrations and debates that got the most reaction in the last 30 days. That is pain mining straight up: you write your next reel on what people actually ask, not on what you imagine.
Run competitive research. With --competitors, you get in one command what people are saying about a competitor and its alternatives, ranked by engagement. What they love, what they complain about, what just shipped.
Prep a call or a meeting. You type the name of a person or a company, and the engine aggregates their X posts, Reddit threads, videos and GitHub activity into one brief. You walk into the meeting knowing exactly where they stand.
Read market signals. Job postings and careers pages become clues: a company hiring in security or infrastructure tells you where it is investing. You anticipate instead of reacting.
I use it mostly for two things: pulling content angles on AI and marketing, and keeping an eye on what is moving on the tooling side. One command replaces an hour of scrolling, and I walk away with sources I can cite.
Before you start
Three reflexes to get going fast and clean.
--emit=html, you get a clean, dark-mode file you can send to a client or keep in your notes.last30days is your market research in one command: around fifteen platforms queried at once, results ranked by real engagement instead of SEO, a sourced brief saved locally. Install the plugin, run your first search on the free sources, and keep the API keys for when you want to push further.
Want to go further?
In the Lab, I share the Claude resources I actually use and how I wire them into my systems.
A dedicated session or program, tailored to your tools and use cases.
And day-to-day, I post one reel a day on Instagram: @quentin_iamarketing