Claude has been having a big moment lately, and it’s not just hype. It’s genuinely one of the most useful AI tools I’ve come across. It’s long stopped being just a tool I open on occasion - I now practically live in it. It’s there when I need ideas for the novel I’m chipping away at, references for design projects, random research I actually need to retain, literally everything that I’d normally spend four times as long Googling or procrastinating on. It became my default first stop for a weird spread of things.
Projects is a big reason why. It’s the feature that lets Claude actually remember things between conversations - your context, documents, your whole setup - so you’re not re-explaining yourself every single time you open a chat. And once I set it up, a lot of the other stuff I used to manage separately just starts to feel like extra work for no reason. Projects isn’t exactly a hidden gem, but for something that does this much heavy lifting it’s surprisingly under-talked about.
What Claude’s Projects feature actually is
It’s not just a folder for your chats
The short version: every normal Claude chat starts completely fresh - it doesn’t know what you’re working on, what you’ve already covered, or anything about you. Projects change that. They’re a persistent workspace where your custom instructions and uploaded documents carry over across every conversation inside that project, so Claude already has the context before you’ve typed anything.
What you’re working with inside a Project is three things. Custom instructions, which are where most of the real value is. Here you give it context about what the project is, what you’re trying to do, how you want responses structured. There’s a knowledge base, where you can upload PDFs, Word docs, CSVs, plain text, images - up to 30MB per file - and it stays referenceable across every chat.
Free users get five projects, which is genuinely usable. Paid plans open up unlimited projects and better RAG handling when the knowledge base gets large - basically smarter retrieval when you’ve loaded a lot in. The five-project free cap will be enough for some people and might be immediately constraining for others.
Setting one up properly takes maybe ten minutes but it’s easy to skip the parts that matter. Name it specifically. Write real instructions that include the actual context: what this project is, what you’re working on, how you want it to respond, what it should keep in mind, what to avoid, and so on. Seed the knowledge base before you start chatting, but it’s also worth revisiting it throughout the project to iterate or expand.
I connected these tools with Claude and my productivity doubled in no time
My journey form overworked to automated.
How Projects replaced half the mess I used to manage
Notes I forgot about, bookmarks I never opened, and tabs for “later”
My notes situation was bad in a specific way - I would enthusiastically create them, but then forget where I put them due to using multiple note-taking tools, or that they even existed. Half of it was ideas I captured in the moment and never went back to, which sort of defeats the purpose. The more notes I accumulated, the less I trusted any of them to actually be useful.
What Projects replaced wasn’t the notes themselves, it was the habit that created them. Now when I have a half-formed idea or a question about, say, the novel I’m working on, I open that Project and work through it in a conversation. The thinking happens there instead of in a note that’ll make no sense in a week. The conversation is the note, and it has context around it so it’s actually readable later.
Bookmarks went the same way. I have hundreds, maybe more, most of them from a version of me who was very optimistic about how much reading they were going to do later. Anything genuinely useful now goes into the Project knowledge base instead - either the full content pasted in as text, or the URL as a weblink reference. Now it’s actually there when I need it, but I don’t have to make the deliberate action of looking for and opening the page every time.
Pinned and opened browser tabs are basically the same problem as bookmarks, just more visible. Each one is something that felt too important to close. The habit shift was pretty simple: anything I’d have opened a new tab to "remember", I put into the relevant Project instead. So now, whenever I get, say, design inspiration, all of it goes into my weekly Design Project in one swoop. Next time I want to tackle the design project, I don’t have to worry about lost references.
A simple trick I use is taking screenshots. Whether it's of a note or of design inspiration, and adding it as a reference file to my Project. Claude can read and interpret all images, so it will be able to see exactly what you're working with, and screenshots are also much quicker than manual copy-pasting.
Claude doesn't replace my PM tools but it fills the gaps between them
Instead of swapping out my PM tools, Claude fills the messy gaps where context and follow-through usually break down.
What Projects won’t do
Honest caveats before you ditch all your other tools
Projects is not a notes app, and if your system is built around something like Obsidian with backlinks and a knowledge graph, it won’t replace that. There’s no full-text search across conversations and no networked notes. The knowledge base is also a snapshot, not a live feed. Whatever you upload is static - if the document changes, you have to reupload it manually.
Context limits are worth knowing about too, especially on the free tier. If you load the knowledge base heavily you can hit the token ceiling quicker. Paid users get automatic RAG handling when it fills up, which basically means smarter retrieval as the Project gets larger. Free users can manage it by periodically swapping out older or less relevant docs.
The five-project cap on the free tier is the other one. It feels fine at first, but if you’re running a few active things at once you’ll hit it faster than expected and have to make some decisions about what stays. Lastly, it's worth noting that while you can add weblinks in the text field as references, Claude doesn't index the page at setup. When chatting, it'll try the URL directly, fail or succeed, and either way usually seems to end up pulling the right content via search.
Just give it a shot
Projects is one of those features that’s hard to explain why it’s good until you’ve actually set one up properly and lived in it for a couple of weeks. The value isn’t obvious from the outside. But if you’re someone who loses things in their own notes, or has a bookmark folder that’s more anxiety than organization - it’s worth setting up one Project, loading it properly, and seeing how different a conversation feels when Claude already knows what you’re doing.
