last30days is an open-source skill for Claude Code and other AI agents. You type a topic; it searches what was said about it in the last 30 days on Reddit, Hacker News, GitHub, YouTube or X, ranks results by real engagement, then the AI writes a sourced summary. The basic setup is free, with no keys.
Tools
Claude Code, last30days skill, Python 3.12+
Price
Free basics, optional paid sources
Time
10 min to install
Level
Terminal, beginner
A prospect asks what you think of the tool they just bought, or you are torn between two pieces of software for a client. You open ten Reddit tabs and three YouTube videos, and an hour later you still have no overall picture. last30days does the digging for you, and your AI agent turns it into a summary note, with the links.
Where this skill comes from, and what it does
The official repository is mvanhorn/last30days-skill, published by Matt Van Horn under the MIT licence and very active (version 3.28.0 of 11/10/2026, with many contributors). Be careful: copies circulate under other GitHub accounts, so install from this one.
How it works, according to the README:
the agent first works out where the topic is discussed (subreddits, accounts, channels, GitHub repositories);
it queries every source in parallel, including top comments and YouTube video transcripts;
it scores each result by engagement (upvotes, likes, views), relevance and freshness, and merges duplicates across platforms;
your AI agent writes the summary, then answers your questions about what it just read.
Research is saved in ~/Documents/Last30Days/. The README also states "no tracking, no analytics".
Install last30days in Claude Code
Add the catalogue and the pluginIn a Claude Code session, type one after the other:
For other agents (Codex, Cursor, Gemini CLI and so on), the README gives: npx skills add mvanhorn/last30days-skill -g. Use one method per computer, otherwise the command shows up twice.
Run a first searchOn first run, a setup wizard asks you a few questions. Reddit, Hacker News and GitHub work straight away, with no key.
Add YouTube (free)Transcripts go through yt-dlp. On a Mac:
Terminal
brew install yt-dlp
On Claude.ai (the website), the README suggests downloading the last30days.skill file from the latest release and uploading it under Customize > Skills, after turning on code execution.
Requirements, keys and cost
In Claude Code, your Claude model plans the search and writes the summary: no extra AI key is needed, but each search uses your plan. For sources, according to the README table as of 17/10/2026:
Reddit (with comments), Hacker News, GitHub: free, nothing to set up.
YouTube: free with yt-dlp.
X (Twitter): either the session already open in your browser (free), or an official X API key, paid in credits on your developer account, or a key from another provider.
TikTok, Instagram, Threads, LinkedIn: a ScrapeCreators key, 10,000 free calls then pay as you go.
Web search: a Brave Search key, 2,000 free queries a month.
Keys go in ~/.config/last30days/.env, or in the macOS Keychain. Never in a shared project folder.
Example: prepare a tool choice for a client
A client is torn between two automation tools. Before the meeting:
Claude Code
/last30days n8n vs Make
Then, in the same session, once the summary is on screen:
Prompt
Based on this research, write a one-page note for a non-technical client:
1. What users like and dislike about each tool this month, with the link to each source.
2. Recent problems (outages, price changes) reported by several people.
3. What remains uncertain or rests on a single account.
Do not make up anything that is not in the sources.
To share the result as is, the README documents a self-contained HTML page output (--emit=html).
Useful settings
Summary language: sources are mostly in English and so is the summary. If your client works in another language, ask for the note in that language.
Videos in other languages: by default, YouTube captions are fetched in English, Spanish and Portuguese. Add this line to ~/.config/last30days/.env to include French, for example:
.env
LAST30DAYS_YT_SUB_LANGS=en,fr
Diagnostics: ask for /last30days doctor to find out which source is missing and why, without showing your keys.
Limits
The result tells you what people say, not what is true. A highly upvoted Reddit thread can be wrong: check the important points at the source.
Each platform can change its access overnight. Through the official API, X only covers roughly the last week, unless you have extended access.
A search takes several minutes and uses your plan: keep it for decisions that matter.
Use it for topics, products and companies, not to profile a private individual.
Watch out for browser cookies. For X, the wizard offers to read your browser's login cookies, in other words to reuse your session. It asks for your consent, and saying no does not stop the rest from working. The repository itself acknowledges that this kind of scraping puts the account used at some risk. My advice: say no on your main browser, or use an API key. Before running anything, the script's --preflight option shows what it will read and write, without running a search. And as with any skill, read the SKILL.md before installing.
Want this research in your inbox every Monday, without opening a terminal? That is an automation I can set up.
First step
Let's talk about what eats up your time
Thirty minutes on a video call. You walk me through how your business runs, and I tell you what could be simpler, roughly what it would cost, and what I would leave alone if I were you.