What is last30days?
last30days is a slash-command research skill for Claude Code, Codex and about 50 other agent hosts. Given a person, company or topic, it fans out across platforms Google does not rank well, scores each result by upvotes, likes, views and prediction-market money, and has the agent synthesise one cited brief on the last 30 days.
Reddit, Hacker News, Polymarket and GitHub work with no key. X, YouTube, TikTok, arXiv, Techmeme and a long list of others switch on when you add API keys or browser sessions of your own. Matt van Horn maintains it; the engine is a Python script the skill calls, research runs on your machine, and the project states it has no tracking or analytics.
Is last30days worth it?
The idea is sharper than the usual "AI search" pitch. Each platform is a walled garden with its own API and auth, so no single model can see across all of them; last30days bridges a dozen with your own keys and browser sessions, then ranks what it finds by real engagement rather than SEO. An earlier write-up describes the weighting as text relevance 35%, engagement 25%, source authority 20%, cross-platform convergence 10% and recency 10%. The output is a brief with upvote counts, view counts and Polymarket odds attached, which is genuinely hard to assemble by hand.
Free sources carry it a long way. Reddit, Hacker News, Polymarket and GitHub need no credentials; the Reddit path reads RSS and scrapes the public site for comment scores. Everything richer, X posts, YouTube transcripts, TikTok engagement, is behind a key or a logged-in session you provide.
The catch: those sessions are the risk. The cleanest route sets the official X API bearer token, but the skill also supports reading your browser cookies to call X's internal endpoints, and one fork adds a fixed proxy to keep an IP consistent with the session, which is the sort of thing platform anti-fraud rules exist to stop. Reddit's free path is scraping. You are responsible for staying inside each platform's terms, and the convenient options do not. It also holds whatever credentials you add, and quality depends on the reranking model you point it at.
The skill is active and widely used: 67,668 stars, commits into November 2026, and a community security pass that fixed stored-XSS in the HTML report and locked down cookie temp files. There is no published accuracy measurement, so treat the synthesis as a strong starting point, not a verified one.
Our take: it is a brilliant pre-meeting and pre-launch briefing tool if you wire up the official APIs; lean on the browser-cookie shortcuts and you are trading that convenience against your accounts.
Where it earns its place
A founder prepping for a sales call
Know what a company and its leaders have actually done in the last month
Run `/last30days <company>` and it pulls recent posts, hiring pages, patent filings and the share-price move into one brief. Free sources cover a lot; add an X or YouTube key for the social layer. Check anything load-bearing against the source.
A content creator tracking a fast-moving field
Find what the Reddit and X community figured out before it reaches training data
Ask `/last30days what's exploding in AI agents?` and the skill switches to discovery mode, sweeping category listings and returning velocity-ranked topics with cross-source numbers. One user put the time saved at about 90 minutes a piece.
An analyst comparing tools or products
Get a side-by-side read backed by live engagement and star counts
`/last30days A vs B vs C` pulls live GitHub stars and community sentiment into a comparison table. The ranking favours what people engaged with, so weight it against your own criteria rather than taking it whole.
Before you commit
- Setup
- Claude Code: `/plugin marketplace add mvanhorn/last30days-skill` then `/plugin install last30days`; other hosts use `npx skills add mvanhorn/last30days-skill -g`. Zero config for the free sources; a setup wizard adds keys. The engine needs Python 3.12+.
- Integrations
- Claude Code, Codex, Cursor, Gemini CLI, Grok Build, OpenClaw, Claude Desktop (as an MCP bundle) and 50+ Agent Skills hosts; sources include Reddit, X, YouTube, TikTok, Hacker News, Polymarket, GitHub, arXiv, Techmeme, Yahoo Finance and more.
- Data and privacy
- Research runs on your machine; the project states no tracking or analytics. Keys live in a local `.env` or the OS keychain. Reranking is opt-in: with no reasoning key it scores locally and sends nothing out; with one, topic and snippets go to that provider.
- Cost at scale
- The skill is free; cost is the provider keys and the reranking model you enable, plus any paid scraping API (such as ScrapeCreators or Bright Data) you opt into for TikTok, Instagram or Amazon.
- Limits
- Browser-session and scraping routes may breach platform terms; richer sources need your own keys; no published accuracy figures; results vary with the reranking model.
What’s good
- Searches a dozen platforms in parallel and ranks by real engagement
- Reddit, HN, Polymarket and GitHub work with no API key
- Cited brief with upvote, view and prediction-market numbers
- Runs locally with no tracking; active community security work
What’s not
- Browser-cookie and scraping routes can breach platform terms
- Holds whatever provider keys and sessions you add
- Quality depends on the reranking model you supply
- No published accuracy measurement
Best for
Researchers, founders and marketers who want a fast, engagement-ranked read on a person or topic and will wire up official platform APIs.
last30days alternatives
last30days: common questions
- Is last30days worth it?
- last30days does the half-hour of Reddit and X trawling you would otherwise do before a meeting, as long as you are comfortable pointing a scraper at platforms through your own logins. We rate it 3.8 out of 5.
- How much does last30days cost?
- Free and open source (MIT). You supply, and pay for, any provider keys and reranking model you turn on.
- What are the downsides of last30days?
- Browser-cookie and scraping routes can breach platform terms; holds whatever provider keys and sessions you add; quality depends on the reranking model you supply; no published accuracy measurement.
- Who is last30days best for?
- Researchers, founders and marketers who want a fast, engagement-ranked read on a person or topic and will wire up official platform APIs.
- How much does last30days cost for a team or in production?
- The skill is free; cost is the provider keys and the reranking model you enable, plus any paid scraping API (such as ScrapeCreators or Bright Data) you opt into for TikTok, Instagram or Amazon.
- Is last30days safe for company data?
- Research runs on your machine; the project states no tracking or analytics. Keys live in a local `.env` or the OS keychain. Reranking is opt-in: with no reasoning key it scores locally and sends nothing out; with one, topic and snippets go to that provider.
- What are the best last30days alternatives?
- The closest alternatives we have reviewed are GPT Researcher and OpenResearch.
Answers checked Nov 14, 2026.