Raindrop.io "Stella": AI librarian for the bookmark collection

The AI of the Bookmark Manager searches for links, summarizes articles, or helps clean up bookmarks.

listen Print view
2 min. read

Users of the online bookmark manager Raindrop.io have recently found an inconspicuous button labeled “Questions” in the user interface. Behind it is the assistant Stella. She is intended to transform the bookmark collection into a searchable memory much better than a full-text search. And it works surprisingly well. If you ask, for example, “Please find all foundational articles on AI,” she replies with the most relevant hits, as shown in the screenshot.

The bookmark manager now understands natural language questions and returns commented link collections.

Stella can also summarize saved articles, explain complex content in simple language, and help with cleanup – for example, finding duplicates, merging tags, or moving unsorted bookmarks into appropriate collections. Before she makes any changes to the library, she politely asks for permission.

Videos by heise

The smart assistant uses OpenAI's open-source GPT OSS 120B, which is hosted on its infrastructure, according to the operator. User data does not leave the servers and is not used for training. For her work, the assistant accesses the static copies that the service creates from the content of each saved link. It extracts up to 300,000 characters, also taking PDF files into account.

Thanks to the MCP server, Stella can also be used from other applications, such as from Claude here.

Stella is currently in the beta phase and is available as early access exclusively for Pro subscribers (starting at around $28 per year) in the web app. In addition to English, Spanish, Chinese, Japanese, and Russian are supported, among others – German is not on the official list, but it worked smoothly in our tests. Additionally, Raindrop.io offers an MCP interface, allowing the bookmark library to be used directly from ChatGPT or Claude.

(jo)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.