last30days: research any topic from the last 30 days
Added NewCopyable prompt below
Published Last verified Released
Our verdict
Web search tools mostly return articles and blog posts. last30days looks at what people are actually discussing: Reddit threads with their upvotes, Hacker News points, X likes, YouTube transcripts and Polymarket odds, all limited to the last 30 days, and your agent turns the results into a brief with sources. Reddit, Hacker News, Polymarket, GitHub, patent filings and careers pages work with no API key. X, YouTube, TikTok and others are opt-in, with keys or a browser login. Use it for pre-meeting briefs, tool comparisons ("X vs Y vs Z") and spotting topics while they are still rising. Treat the brief as a starting map, not a verdict: engagement shows what is popular, not what is correct, so open the linked threads before you quote them.
- Who it's for
- Builders, founders, content creators and researchers who need a quick, sourced read on what people are saying about a tool, company, person or topic right now
- Leverage
- Turns an hour of tab-hopping across Reddit, HN, X and YouTube into a few minutes and one sourced brief, with no paid search API needed to start
- Effort
- Low
- Date added
What it does
last30days (mvanhorn/last30days-skill, MIT, Python 3.12+) is an Agent Skills package with a Python engine. A /last30days <topic> run searches the sources available to it in parallel, keeps results from the last 30 days, ranks them by real engagement (upvotes, likes, views, points, market volume) and hands them to your agent, which writes a sourced brief. Comparison topics get a side-by-side table, and a question like "what's exploding in A.I. agents?" switches it to discovery mode. Free with no key: Reddit (with comments), Hacker News, Polymarket, GitHub, StockTwits, and US patent filings and careers pages for company topics. arXiv, Techmeme and Yahoo Finance use free CLIs that first-run setup installs. Opt-in: X (official API bearer token, xAI key, or a logged-in browser session), YouTube (yt-dlp), Bluesky, a ScrapeCreators key for TikTok, Instagram, Threads, Pinterest and LinkedIn (10,000 free calls, then pay as you go), Brave and Perplexity. The project says research stays on your machine except for the search APIs you call. If you set a model key for reranking, your topic and snippets go to that provider. Optional local SQLite store with LAST30DAYS_STORE=1. Risks: browser-cookie reads and "Reddit as you" use your logged-in sessions, so use --preflight to see what a run would read and write, and --no-browser-cookies to turn cookie reads off. Setup also installs extra CLIs. The project ships often (v4.3.2 on 17 Nov 2026), so your installed version may behave differently from the README.
How builders use it
- In Claude Code, run /plugin marketplace add mvanhorn/last30days-skill, then /plugin install last30days. In Codex, Cursor, Copilot or Gemini CLI, run npx skills add mvanhorn/last30days-skill -g.
- Start with the keyless sources: /last30days <your topic>. Add X or YouTube later through the setup wizard only if you need them.
- Before a sales call or partner meeting, run it on the company or person and paste the brief into your notes.
- Compare tools before you pick one: /last30days Cursor vs Claude Code vs Codex. Then open the top threads it cites.
What to expect
A dated, sourced brief on what real users are saying, ready for a meeting, a buying decision or a content angle, in minutes instead of an afternoon of searching.
Copy this: Install + first runs
# Official source: github.com/mvanhorn/last30days-skill (MIT). Needs Python 3.12+.
# Claude Code
/plugin marketplace add mvanhorn/last30days-skill
/plugin install last30days
# Codex, Cursor, Copilot, Gemini CLI and other Agent Skills hosts
npx skills add mvanhorn/last30days-skill -g
# Keyless first run (Reddit, HN, Polymarket, GitHub)
/last30days code review agents
# Compare tools
/last30days Cursor vs Claude Code vs Codex
# Discovery mode
/last30days what's exploding in coding agents?
# Safety: see what a run would read and write first; turn off cookie reads
# (run from the skill directory) scripts/last30days.py --preflight
# add --no-browser-cookies to a runRelated discoveries
See the newest picks →- ToolResearch & Automation
imozuru: research an unfamiliar field down to the original sources
An MIT Claude Code skill that maps a new field, follows citations to the original sources and cites the exact lines.
NewEffort: MediumPrompt#Plugin / extension#Open source
- ToolResearch & Automation
agent-desktop: computer use for agents via the accessibility tree
An Apache-2.0 Rust CLI and C library from Vercel Labs that gives any agent structured, retry-safe control of desktop apps through the operating system's accessibility tree.
Effort: LowPrompt#Open source#CLI
- ToolResearch & Automation
Open Deep Research (LangChain)
LangChain's MIT deep research agent on LangGraph: plans, researches with parallel sub-agents and writes a report, with the models and search you choose.
Effort: High#Open source
- ToolResearch & Automation
GPT Researcher
An open-source agent that plans research questions, searches the web or your documents and writes a cited report, with the model you choose.
Effort: Medium#Open source
Get the best new A.I. picks in your inbox
The best new agent workflows and tools we’ve vetted, with a clear verdict on each. Short, occasional, useful.