Nosey, a DIY RSS scraper agent

The AI that pokes into your feeds so you don't have to.

Let’s be honest — RSS feeds are amazing in theory, but in practice? Total information overload.

You spend ten minutes adding a shiny new batch of tech blogs to your email sub list, feel like a productivity god for half a day, and then... life happens. Three days later, you open your mail to "999+ unread posts."

At that point, checking the news feels less like staying informed and more like doing unpaid data entry. Half the articles are carbon-copy rewrites of the exact same press release, three are sponsored posts for ergonomic mousepads, and maybe two are actually worth your time.

Nosey is a lightweight, zero-cost, set-and-forget tool that tames your *RSS * firehose. It automatically fetches your feeds, weeds out the duplicate noise, picks the actual gems using AI, and delivers a clean daily digest right to your repo (or your inbox).

No monthly SaaS subscriptions. No bloated web apps. Just a cozy little background process doing the heavy lifting while you sleep.

What makes Nosey tick?

Instead of dumping everything into one monolithic, fragile script that crashes the second an XML feed changes format, Nosey uses small, modular agents.

Think of each agent as a tiny, single-purpose worker bot. Here’s what it does every day:

1. Fetches the latest posts from whatever RSS/Atom links you give it.
2. Kills duplicate coverage across different sites (using Jaccard similarity — because reading the exact same tech acquisition headline rewritten five times is a special kind of developer torture).
3. Hand-picks the top 8–15 stories using Google Gemini 2.5 Flash based on topics you actually care about.
4. Writes a sleek Markdown digest and archives it safely in your repository.
5. (Optional) Sends a clean email straight to your inbox via Resend.

All the agents share the same underlying core engine. That means you never have to write complex scraping or AI parsing logic — you just edit a simple configuration file and let it run.

Pre-built agents ready to roll

You don't need to build anything from scratch. Out of the box, Nosey comes with five ready-to-use agents:

Want to tweak what they look for? Or spin up a custom "Mechanical Keyboards & Indie Games" agent? It takes about two minutes.

Truly zero cost ($0 / month forever)

We’ve all built side projects that were supposed to be "free" until a serverless function went rogue over the weekend and hit us with a $40 AWS surprise.

Nosey is built specifically to sit comfortably inside generous free tiers:

No hidden API paywalls, no "trial expires in 14 days" banners, no credit card required.

How to get it running in under 5 minutes

You don't need to deploy servers, configure Docker containers, or manage database clusters.

1. Fork or clone the Nosey repository.
2. Grab a free Gemini API key from Google AI Studio (takes 30 seconds).
3. Save your secrets in your GitHub repo (Settings ➔ Secrets and variables ➔ Actions):
   • GEMINI_API_KEY (Required)
   • RESEND_API_KEY & EMAIL_TO (Optional — only if you want email digests)
4. Push your code to GitHub.
5. Click "Run workflow" under the Actions tab to test your first run.

Boom. In about 60 seconds, your first compiled digest will appear in the digests/ folder.

By default, GitHub Actions will automatically run Nosey every morning at 08:00 UTC so your digest is ready when you start your day.

Customizing your feeds without breaking stuff

Open up any agent file (like index.js). You'll see a straight-forward CONFIG object at the top:

javascript
const CONFIG = {
  FEEDS: [
    '[https://www.theverge.com/rss/index.xml](https://www.theverge.com/rss/index.xml)',
    '[https://techcrunch.com/feed/](https://techcrunch.com/feed/)'
  ],
  EXCLUDE_KEYWORDS: ['sponsored', 'advertorial', 'deal alert'],
  RANK_CRITERIA: [
    'Major open-source software releases',
    'Significant AI breakthroughs',
    'High-impact cybersecurity incidents'
  ],
  MIN_STORIES: 10,
  MAX_STORIES: 15
};