Saving ideas is easy. Using them is where most people lose. Here's how I wired my AI second brain into Claude, ChatGPT & Gemini — so my research works for me.
Awesome! By using Model Context Protocol (MCP) to bridge your research directly into Claude, ChatGPT, and Gemini, Cluing transforms the traditional 'second brain' from a static digital graveyard of saved links into an active context engine for your daily AI workflows.
"Saving things is only half the problem" is where every second brain I've tried quietly died. Saving feels productive, so I'd collect for months and never once go back in. Curious whether the weekly recap actually breaks that for you or just makes the pile easier to look at.
A recap that runs on a calendar mostly just makes the pile easier to look at, it doesn't make you act, so on its own it won't break the collect-and-die loop. What does break it is pulling something up only when a live decision needs it, not on a schedule: the saved thing earns its keep only when a real question reaches back in and puts it to work. We do this directly with the clients we work with when we restructure a stack from loose prompts and skills into automated workflows, and the first cut is always asking what a saved thing has actually saved you and whether that can be shown. If it can, it's a keeper; most of the time it can't, and that's the quiet weight dragging the system down.
The scheduled task definitely helps. But what helps most is the MCP itself, because I can pull up the things I’ve saved whenever I want to focus on something specific, things like
“what should I read this week?” and have it pull from the collection of articles I’ve bookmarked. But mostly, for my work, I use it a lot for research, and that’s usually very project-specific and it's easy for me to pull everything I saved when I need it.
It ultimately depends on what you need most based on what things you save and what you need them for afterwards.
The markdown-with-metadata question is the one to keep asking, and it matters more than portability. A second brain only pays off when the AI can pull the right note when you need it, and a flat PDF export drops exactly what makes that possible: the tags, headings, and links that let it find the slice instead of re-reading everything.
So the real test of these tools isn't which apps it replaces, it's whether your knowledge stays machine-readable after it leaves the app. If a tool only exports a rendered document, your notes stay readable but stop being searchable by the AI you saved them for.
Cluing is amazing! I’ve recommended it to my friends who struggle with keeping things in one coherent place. I’m all over the place too with 15 open tabs, Google Drive and Notion so tools like that are super helpful. I love the prioritization list, Daria.
Love hearing that, Anna! That prioritization list is so needed, given how many things I save as "things I should try" 😂 At one point a couple of months ago I was also thinking about building an extension for saving testimonials because whenever I needed them for my About page or landing pages, they were always so hard to find. But now, yeah, I can use Cluing for so many different things.
I’ve implemented your os system . I’m curious how this interacts with that system and where the touch points are I.e what lives in the file system of that vs cluing and what decides. Have you written about this?
Hey Phil, it depends on how you're using Cluing. For example, a scheduled task that pulls information from Cluing could be built into the OS, so the things you've saved across apps automatically ends up in the right workspace and folder. But for the most part, it depends on how you use it. You decide whether that context is relevant to the OS and what kinds of information should be transferred into it, so you can also teach the system how to do that for you.
So, you make cluing as your primary dashboard then claude as second? i tend to find another one centralize dashboard since notion n clickup has pros cons on collecting ideas-things, thx for your insight
Not sure I follow what you mean. Claude isn't really a dashboard, so Cluing's interface is what helps you see all your saved things in one place. Claude is how I pull them in to work on them or push more saved things from my Claude work back into Cluing, so everything sits together into Cluing’s dashboard.
I’ve tried using it but couldn’t stick with it. It needs quite a bit of maintenance and can get cluttered (and overwhelming) pretty quickly. For AI context for example I built my own setup using .md files stored in GitHub and Drive (and they only contain what AI needs so it works properly). Cluing is much easier to set up, no learning curve, and doesn't require that much upfront effort (+technical skill) to make it work.
Abha, the reason Obsidian tends to buckle as AI context isn't only the upkeep, it's that a vault is built for a human to browse and a model context is built to be read cold, and those two want opposite things. Your brain rewards a rich, linked, slightly messy vault; a model rewards ruthless pruning to just the facts a task needs, or it drags in noise and drifts. That's why Daria's .md-files-with-only-what-AI-needs setup holds: it's a context store, not a knowledge garden. Pick the structure for the reader you're actually optimizing for, and most of the maintenance pain goes with it.
This is a great walkthrough of wiring a second brain into the tools you actually use daily. I took a similar route but built my whole personal OS - apps, notes, artifacts - around an AI agent that manages it end to end. Wrote it up here if useful: creatism.substack.com/p/thinking-of-building-a-personal-os
Will def try it!
can’t wait to hear what you think of it once you do!
Awesome! By using Model Context Protocol (MCP) to bridge your research directly into Claude, ChatGPT, and Gemini, Cluing transforms the traditional 'second brain' from a static digital graveyard of saved links into an active context engine for your daily AI workflows.
"Saving things is only half the problem" is where every second brain I've tried quietly died. Saving feels productive, so I'd collect for months and never once go back in. Curious whether the weekly recap actually breaks that for you or just makes the pile easier to look at.
A recap that runs on a calendar mostly just makes the pile easier to look at, it doesn't make you act, so on its own it won't break the collect-and-die loop. What does break it is pulling something up only when a live decision needs it, not on a schedule: the saved thing earns its keep only when a real question reaches back in and puts it to work. We do this directly with the clients we work with when we restructure a stack from loose prompts and skills into automated workflows, and the first cut is always asking what a saved thing has actually saved you and whether that can be shown. If it can, it's a keeper; most of the time it can't, and that's the quiet weight dragging the system down.
The scheduled task definitely helps. But what helps most is the MCP itself, because I can pull up the things I’ve saved whenever I want to focus on something specific, things like
“what should I read this week?” and have it pull from the collection of articles I’ve bookmarked. But mostly, for my work, I use it a lot for research, and that’s usually very project-specific and it's easy for me to pull everything I saved when I need it.
It ultimately depends on what you need most based on what things you save and what you need them for afterwards.
Cluing sounds amazing. Can it export knowledge—preferably in Markdown with metadata intact? What apps do you consider it a replacement for?
Everything you save into topics can be exported either as a PDF or as a csv file. We'll be adding markdown in the next few sprints as an option too 😄
The markdown-with-metadata question is the one to keep asking, and it matters more than portability. A second brain only pays off when the AI can pull the right note when you need it, and a flat PDF export drops exactly what makes that possible: the tags, headings, and links that let it find the slice instead of re-reading everything.
So the real test of these tools isn't which apps it replaces, it's whether your knowledge stays machine-readable after it leaves the app. If a tool only exports a rendered document, your notes stay readable but stop being searchable by the AI you saved them for.
Cluing is amazing! I’ve recommended it to my friends who struggle with keeping things in one coherent place. I’m all over the place too with 15 open tabs, Google Drive and Notion so tools like that are super helpful. I love the prioritization list, Daria.
Love hearing that, Anna! That prioritization list is so needed, given how many things I save as "things I should try" 😂 At one point a couple of months ago I was also thinking about building an extension for saving testimonials because whenever I needed them for my About page or landing pages, they were always so hard to find. But now, yeah, I can use Cluing for so many different things.
Also, thank YOU for spreading the word! 💜
I’ve implemented your os system . I’m curious how this interacts with that system and where the touch points are I.e what lives in the file system of that vs cluing and what decides. Have you written about this?
Hey Phil, it depends on how you're using Cluing. For example, a scheduled task that pulls information from Cluing could be built into the OS, so the things you've saved across apps automatically ends up in the right workspace and folder. But for the most part, it depends on how you use it. You decide whether that context is relevant to the OS and what kinds of information should be transferred into it, so you can also teach the system how to do that for you.
I hope I answered your question?
thank you this helps!
So, you make cluing as your primary dashboard then claude as second? i tend to find another one centralize dashboard since notion n clickup has pros cons on collecting ideas-things, thx for your insight
Not sure I follow what you mean. Claude isn't really a dashboard, so Cluing's interface is what helps you see all your saved things in one place. Claude is how I pull them in to work on them or push more saved things from my Claude work back into Cluing, so everything sits together into Cluing’s dashboard.
Hi!! I've been building this in Obsidian. Have you tried that? I'd never heard of Cluing before...
I’ve tried using it but couldn’t stick with it. It needs quite a bit of maintenance and can get cluttered (and overwhelming) pretty quickly. For AI context for example I built my own setup using .md files stored in GitHub and Drive (and they only contain what AI needs so it works properly). Cluing is much easier to set up, no learning curve, and doesn't require that much upfront effort (+technical skill) to make it work.
How's your experience so far with Obsidian?
Struggling! That's why I asked...will check out Cluing. Thanks Daria!
Abha, the reason Obsidian tends to buckle as AI context isn't only the upkeep, it's that a vault is built for a human to browse and a model context is built to be read cold, and those two want opposite things. Your brain rewards a rich, linked, slightly messy vault; a model rewards ruthless pruning to just the facts a task needs, or it drags in noise and drifts. That's why Daria's .md-files-with-only-what-AI-needs setup holds: it's a context store, not a knowledge garden. Pick the structure for the reader you're actually optimizing for, and most of the maintenance pain goes with it.
Got it, thank you!
Meta just approved us, so Instagram and Facebook integrations for mobile are coming next week 🎊
🎉🎉
This is a great walkthrough of wiring a second brain into the tools you actually use daily. I took a similar route but built my whole personal OS - apps, notes, artifacts - around an AI agent that manages it end to end. Wrote it up here if useful: creatism.substack.com/p/thinking-of-building-a-personal-os
Finding the UX frustrating right now.
hey Martin, could you share a bit more context on what you’re expecting & how you’d like the UX to work?