Published: 2026-06-10

Last 30 Days: Open-Source AI Agent That Searches Reddit, X, and Polymarket

Chapters / key moments (click to jump — plays here on the page)

Income Stream Surfers covers Last 30 Days — an MIT-licensed open-source AI agent skill with ~40K GitHub stars. Unlike Google (which ranks by editorial authority and SEO), Last 30 Days searches Reddit, X, YouTube, Hacker News, Polymarket, and GitHub in parallel, scoring results by real engagement: upvotes, likes, and real money (Polymarket odds). It installs on any agent platform as a skill and is positioned as a Perplexity alternative at API cost instead of grounding API cost.

Source video

"This Open Source AI Agent Searches Reddit, X & Polymarket (FREE)" by Income Stream Surfers — Watch on YouTube →

Key Takeaways

MIT license, ~40K GitHub stars. Installs as a skill on any agent platform (Claude Code, Hermes, etc.) — not a standalone app.
Searches 7+ platforms in parallel: Reddit, X (Twitter), YouTube, Hacker News, Polymarket, GitHub, and the web — filtering to results from the last 30 days only.
Scored by real engagement, not editors: Reddit upvotes, X likes, Polymarket odds backed by real money, YouTube views. Cuts through SEO-optimized evergreen content.
Bring your own API keys and browser sessions to unlock walled-garden platforms. Web search uses Brave API (2,000 free requests/month).
Uses Claude, Gemini, Google, and ChatGPT for synthesis — bring your own keys for multi-model judgment.
Cheaper than Google grounding for high-volume use cases; not entirely free (Scrape/auth keys required for some platforms).
Best use cases: pre-meeting research on a person or topic, competitor analysis, trending content for channels, finding current prompts, news briefings before trips.

How It Works

Each platform is a walled garden with its own API tokens and authentication. Last 30 Days acts as an orchestrator: you give it a topic, it figures out which subreddits, YouTube channels, and hashtags are relevant before making a single API call (a "pre-search brain"), then searches all of them simultaneously. A second LLM judge scores every result for relevance, synthesizes the top signals into a single HTML brief.

The creator notes this is "social relevancy, not SEO relevancy" — it surfaces what people are actually engaging with right now, which often differs substantially from what ranks on Google for the same query.

Caveats

Not entirely free — Reddit scraping and some platform access require API keys or browser session auth
Reddit scraping may conflict with Reddit's terms of service at scale
Brave free tier caps at 2,000 web searches/month — not enough for production-volume usage
As AI content dominates social platforms, "social relevancy" searches increasingly surface AI-generated content about AI

Related Guides