We saved bookmarks for years. Hundreds of them. Tutorials we swore we would read, tools we planned to try, articles we wanted to reference "someday." They piled up in random folders, in the Bookmarks Bar, in no folder at all. And we never went back to organize them.
We are Miguel and Max, two developers who share the same bad habit. At some point, we stopped pretending we would fix it by hand. Instead, we decided to build something that would fix it for us.
That is how MarkMind was born.
The Idea Behind MarkMind
The concept was straightforward. What if AI could look at the page you are saving, understand what it is about, and place it in the right folder? Not a new system to learn. Not a separate app. Just your bookmarks, organized the way they should have been all along.
MarkMind is a Chrome extension that does exactly that. You visit a page, click the icon, and the AI reads the title, the URL, and the page content. It looks at your existing folder structure and suggests where this bookmark belongs. You review the suggestion, approve it or change it, and you are done.
No dragging. No "New Folder." No guilt about the mess you have been ignoring.
From Idea to Product: The Journey
We started building in January 2026. The first version was simple. A Chrome popup that could send page data to Google's Gemini API and get a folder suggestion back. It worked, but it was limited.
Then the real questions started coming in.
What if people do not use Gemini?
Not everyone has a Gemini API key. Some developers already use OpenAI. Others prefer Anthropic or OpenRouter. So we added all four providers. You pick the one you already have access to, choose the exact model you want, and you are good to go.
What about the bookmarks you already have?
Organizing one page at a time is great, but what about the 500 bookmarks already sitting in your browser? That became the biggest feature we built. MarkMind can scan your entire bookmark library, let you select which ones need sorting, and then send them to the AI in batches of 25. The AI proposes a new folder structure. You review every suggestion before anything moves.
This feature alone was 38 files and roughly 2,500 lines of code. A Chrome service worker handles the batch processing in the background, so you can close the popup and come back when it is done.
What about dark mode?
We added full dark mode support with automatic system preference detection. Then we spent more time than we would like to admit fixing a background flash that appeared for a split second when opening the popup. The small details matter.
The Bugs That Taught Us the Most
We would love to say everything went smoothly, but that would be a lie.
The slash problem. Some people have folders with / in the name, like DNS/DOMAIN. Our code used / as a path separator. You can guess what happened. We had to build a custom path splitter with a negative lookbehind regex to handle this edge case.
The disappearing root folder. Chrome's bookmark root folder IDs are not consistent across profiles or browsers. We hardcoded them at first. That broke on the second test machine. Now everything is discovered dynamically at runtime.
The AI path mismatch. We send the AI a nicely formatted folder tree. The AI sends back a path. Our code escapes / in folder names, but the AI returns unescaped paths. The lookup fails. We added a fallback match, then a suffix match, then went back and improved the prompts to prevent the mismatch in the first place. Three layers of defense for one string comparison.
The 2,000 bookmark crash. One of our first GitHub issues was from a user with over 2,000 bookmarks and 151 folders. The extension crashed. Sending that much data to an AI in one request is too much. That is exactly why we built the batch processing system.
The security lesson. We accidentally shipped a hardcoded API token in the source code. We caught it, rotated the key immediately, and removed it. Then we moved all sensitive data handling to safer patterns, like passing API keys through headers instead of URL parameters. It was a humbling moment, and it made us better developers.
Every one of these bugs forced us to build something more resilient. The product is better because of them.
What We Built in Two Months
Over 33 merged pull requests and two months of focused work, MarkMind grew from a quick experiment into a fully featured tool:
Single-page organization. Visit any page, click the icon, and the AI suggests the best folder for it.
Bulk organization. Scan your entire bookmark library, select the ones that need sorting, review AI suggestions, and apply them all at once.
Four AI providers. Google Gemini, OpenAI, Anthropic, and OpenRouter, with dynamic model selection for each.
Dark mode. Automatic system preference detection with a manual toggle.
Full review control. Nothing moves without your explicit approval. Every suggestion is shown to you first.
Privacy by design. Only page titles and URLs are sent to the AI. No full page content, no personal data, no tracking.
Open source. MIT licensed. Every line of code is available for you to read and verify.
Privacy Is Not a Feature. It Is the Foundation.
Every AI-powered tool faces the same question: "What are you doing with my data?"
Here is our answer. MarkMind does not collect data. Your bookmarks stay in your browser. The only information sent to the AI provider is the page title and URL, just enough context for a smart suggestion. Your API key is stored locally in Chrome's storage, on your machine, under your control.
We do not run a backend server. We do not have a database. We do not track usage. We do not have analytics.
You bring your own API key. You own your data. And you can audit our code anytime you want.
What Real Users Are Saying
The first issue on our GitHub repository opened with this message:
"This is an amazing extension, it works incredibly well, thank you so much for making it and releasing it under an open source license!"
That same user went on to request subfolder selection support. That kind of feedback tells us something important: people are not just trying the extension. They are using it and wanting more from it.
What Is Coming Next
We are actively building and improving MarkMind every week. Here is what is on the roadmap:
-Insights tab with a bookmark health score, so you can see how organized (or chaotic) your bookmarks really are.
-AI-powered bookmark search, so you can find what you saved even when you forget where you put it.
-Context menu integration for right-click organizing directly from any page.
-Better support for massive libraries with thousands of bookmarks.
-More AI providers and models as the ecosystem evolves.
Ready to Finally Organize Your Bookmarks?
You have been meaning to do it for years. We all have. The difference is that now it takes one click instead of an entire afternoon.
MarkMind is free and open source. Install it, connect your AI provider, and watch your bookmarks find their way home.
Install MarkMind Now
Add MarkMind to Chrome (it takes 10 seconds)
Then come see what we are building:
Watch the demo on YouTube to see it in action
Star us on GitHub if you want to follow the journey
Visit markmind.xyz to learn more
If MarkMind helps you, share it with a friend who also has 400 unsorted bookmarks. You know who they are.
Built with care by Miguel Silva and Max Mendes.
Top comments (1)
Such a relatable problem—bookmark chaos is real! Love how you turned it into a practical AI solution instead of ignoring it!
The focus on privacy and user control makes this even more impressive—great build and thoughtful approach!