A Simple but Powerful Bookmark Feature: Routing Prompts to Your Own AI Subscriptions

I recently added a very simple, yet effective feature to pokke, a personal bookmarking service I am developing.

It is a feature that allows users to attach custom URL links to their saved bookmarks. By using simple variables like {{url}}, {{title}}, and {{description}}, the system dynamically generates a link tailored to that specific bookmark.

The most practical use case for this is passing a prompt with the bookmark’s context directly to an AI search engine like Perplexity. For example, you can configure a custom URL like this:

https://www.perplexity.ai/?q=URL%3A%20{{url}}%0Atitle%3A%20{{title}}%0Adescription%3A%20{{description}}%0APlease%20explain%20the%20content%20of%20the%20specified%20URL%20in%20detail.%0AAlso%2C%20please%20suggest%20some%20interesting%20pages%20on%20related%20topics.

URL: {{url}}
title: {{title}}
description: {{description}}
Please explain the content of the specified URL in detail.
Also, please suggest some interesting pages on related topics.

With just one click from a saved bookmark, this instantly hands over the article’s context to Perplexity, asking it to explain the content in detail and suggest related topics. It dramatically shortens the path from “saving for later” to actually “learning from it.”

You might wonder why I didn’t just integrate an LLM API directly into Pokke to summarize articles automatically in the background.

The main advantage of using URL queries instead of backend APIs is cost efficiency through leveraging the user’s existing SaaS subscriptions. By simply passing the query to a service the user is already paying for (like a Perplexity Pro account), I can offer advanced AI capabilities without bearing the heavy LLM API costs on the service side. It is a win-win for both the developer and the user.

However, there is a clear trade-off. Because it relies on external URL routing rather than a backend agent, it cannot be fully automated. The AI won’t summarize the article the moment you save it; it still requires the user’s manual action (a click) to trigger the process.

Despite the slight friction of needing a manual click, this “Bring Your Own SaaS” approach fits perfectly for small-scale or personal projects. It is a lightweight feature that significantly enhances how I interact with the information I collect.