News

RSS filter

Explore RSS Filter’s reading-based recommendations, OPML conversion and self-hosting option, with important feed-privacy and testing limitations.

Name	RSS Filter
Overview	RSS Filter is a recommendation tool that filters existing RSS feeds using the articles a reader opens. It replaces feed URLs with filtered versions and uses text embeddings and machine learning to recommend similar articles. The public interface accepts a feed URL or an OPML export, allowing readers to keep using an RSS client.
Key features & benefits	Recommendations based on the reader’s article history. Embedding-based similarity and machine-learning filtering. Conversion of multiple subscriptions through an OPML file. Filtered feed URLs for use in an RSS reader. Public source repository and self-hosting documentation.
Use cases and applications	Reduce the volume of articles from a busy set of feeds. Find articles similar to those already read while allowing some topic discovery. Manage subscriptions through an OPML export without replacing the RSS reader itself.
Who uses?	Tech enthusiasts Information specialists Content curators News readers
Pricing	No paid subscription tiers are published on the checked landing page. The linked source repository is MIT-licensed and documents self-hosting; hosting and operation may have separate costs. Do not assume an undisclosed premium plan exists.
Tags	RSS feed, filtering, content recommendation, LLM, OPML
App available?	Web interface plus filtered RSS URLs and a self-hostable project; a separate mobile app was not verified.

Limitations and checks

The repository says recommendations begin after some reading history—10 articles by default—and include some random articles for discovery. Filtering can omit relevant items, so retain access to the original feeds. Reading history and feed URLs are part of the workflow; do not submit private or token-bearing feed URLs without understanding the deployment’s data handling. No feed or OPML file was submitted.

Practical example

Hypothetical editorial example: use a small set of public technology feeds, compare the filtered list with the originals, and check whether important topics were omitted. This describes a possible evaluation, not an observed recommendation result.