A small bakery for the open web.
RSS-powered discovery. Scored against slop. Built for humans and the agents working for them.
FeedBagel turns RSS into a discovery engine for the open web. We taste-test feeds, score articles against clickbait and padding, and serve up the keepers.
The web used to be full of weird, brilliant, hand-written stuff. Then it got buried under highly-processed AI slop. No nutrition, no calories, just keywords tuned for SEO. We're trying to pick out the parts still worth reading, and keep them reachable for the humans who want to read them and the agents fetching them on humans' behalf.
Six things on the counter.
One stack of feeds, scored, sorted, and piped out in whatever format you want to consume it.
- Feed directory
- Thousands of RSS feeds, grouped by site, author, and tag, with sparklines so you can see what’s alive. Browse the directory
- Active discovery
- When a story breaks, FeedBagel hunts down other publishers covering it, checks their feeds, and rolls the keepers in. See today’s tray
- Bagel score
- Every article gets graded against clickbait, buzzword stuffing, and content padding. How the score works
- MCP server
- Wire FeedBagel into an AI agent so it pulls answers from hand-picked sources instead of the open slop. Connect an agent
- Webhooks
- A POST every time a feed you follow publishes something new. Slack, Discord, your own pipeline, whatever. Wire it up
Better sources, better answers.
AI agents are only as good as what they read. Point one at the general web and it ends up regurgitating the same SEO-tuned slop everyone else is producing. Point one at FeedBagel and it pulls from publishers running real RSS feeds, which already filters out a lot of the worst of the web.
The bagel score doesn't specifically detect AI-generated content yet, but we want it to. That work needs model training and infrastructure investment we don't yet have. If you care about the open web staying readable, help us build it.
AI works the kitchen.
Summaries, tags, smart translations, source-cited rundowns that link related stories together, and discovery (fanning out from one story to find other publishers covering it). If AI writes for us in future, it's with care, opinion, and citations. The line isn't “no AI.” The line is no slop, whoever made it. No padded SEO keyword soup, no opinion-free content optimised for ranking instead of reading.
The web's native pipe.
RSS is the web's native publishing format. Almost every blog, magazine, indie journal, university lab, and small press still quietly publishes one. Most don't show up on social platforms. Their feed is the most reliable way to find them.
It's also the universal pipe. Same source, same article, whether the reader is a human in a coffee shop or an AI agent answering somebody's question. FeedBagel sits on top of that. RSS is the plumbing. Discovery, scoring, and curation are what we add.
Built by Prototypr.
FeedBagel is built and run by Prototypr. It grew out of years of running an indie publication and wanting somewhere to point readers that wasn't a walled garden. The agent angle came from watching every AI tool quote the same five SEO-junk sites; we'd rather they reach into a curated index instead.
Bug reports, feed requests, sponsorship interest, integration ideas, kind notes, and grumpy ones all welcome: [email protected].
Curated by hand. Served to humans and agents alike.
