Tips for effectively using Projects?

I recently started using the Project feature daily and realized just how incredibly useful it is! Here are the two main ways I use it:

1. I organize all the details of my novel, including the world-building, character profiles, and timeline, within the project. This way, whenever I ask questions, Claude automatically references these documents, saving me from the hassle of forgetting important details.

2. I also compile my personal reflections and reading notes into the project. By doing this, Claude gains a better understanding of me as a person, allowing for deeper and more meaningful conversations.

Additionally, I love using Claude to create product prototypes—specifically in HTML format, fully interactive. This has significantly reduced the time and effort I spend communicating with technical teams!

i use projects. for example, i have a project for my own personal social media posts. i like to use emojis and i ask Claude to do that. i was able to tell him my interests and when necessary, he can add a reference in there. for example, i'm a big fan of costco...so if a costco joke works out, Claude will find a way to tie it in.

i also have a project for my company branding. i uploaded our website copy and ICPs. so now when i want to write up some marketing material, it is consistent with what we are supporting. the tone and content will be very different for my personal posts. you should give it a try!

Personally, I find the feature incredibly useful. I don't know that it still outperforms ChatGPT at this point (it can use multiple canvases now and reference unrelated chats from your history, so the functionality can be similar in practice now) but I've been using it to help organize my thoughts for a novel I'm writing.

I stored all of my world building related documentation and research as files in a project, and then began working in artifacts to get editing help as I write chapters, then storing each chapter as a new file to the project. By this point, it can research from my world docs details to help inform my writing and catch logical plot holes by reviewing previous chapters, or poiyour where I'm deviating from my plot outline, etc., even critique the writing itself from everything from a grammatical to logical standpoint.

It's pretty helpful for coding too, if you're into that. I can drop mota of my code base into a project (and now, even directly from GitHub) and it can write whole files or update existing files with very little guidance. I still don't trust the outputs without direct human review, but it does save a ton of time and leaves me debugging minor errors immediately, instead of having to write all those bugs myself. 😅

Thats pretty cool! PPLX.ai pro has something called spaces. Its a bit the same. But im not using it at the moment. I prefer having one source of truth on github. What I do often is that I build up linked gists. or just point PPLX to the repo Im working on. and it will scan it and go pretty deep. Its a hit and miss sometimes. I guess I burn lots of internal tokens doing this. So they throttle me sometimes I think. Sometimes it works really great, other times not. And sometimes using PPLX deepresearch with this kind of flow can reveal more. I prefer using the Sonnet 3.7 model doing this. With pplx you can sort of pick which model to use. "So basically prompt. How can I upgrade OAuth from 1.1 to 1.2 for this repo: github.com/username/repo" and it will figure it out. Its sometimes better to use pplx than say Cline with agents having access to the same repo. As pplx is better at looking outward to the world. finding related links etc. and then forming a thesis on how to solve the proble. I ill then copy that brief and give it to Cline with Sonnet 3.7 and the agentic flow will solve the problem. I will then let it work for 10 -30 minutes while I go and play basketball or smth.... What a time to be alive 🚀

My use case is a bit niche, but Projects has been amazing. I am writing a collection of short stories, all taking place in a shared universe. In my project knowledge, I add everything related to worldbuilding, in plain text. I created these documents using Claude, too, building locations, lore, and ideas, and then at the end of those conversations, I asked Claude to summarize so I could add to the project documentation. The output is quite remarkable, and after that, I can dump my drafts and tell Claude to point out inconsistencies, contradictions, etc.

Creator of Desktop Commander MCP here that I recommend to use with Claude Desktop.
One thing that bothers me is that I can't update knowledge files from the chat.
One interesting hack that came from our community was to create a physical txt or md file with knowledge and instruct in project prompt to always read that file when new chat in project is created.
And then something like filesystem mcp or desktop commander can be used to update that file when you want to update project knowledge with Claude assistance.

Projects is must honestly i started using the projects feature about a week ago and i realized really fast that i had been doing it all wrong by not trying it out sooner absolutely awesome and one of the reasons i believe claude is such a unique AI provider. A tip i would suggest is giving it as much information as you can to begin with and it will do the rest.

I used projects so much that I vibe coded my own tool that is essentially Claude projects on steroids. (I called it claude on steroids lol)

I would keep them as niche as possible, one for each workflow, basically then you don't dilute the system prompt too much.

I made a bunch of videos on how I use it haha:
https://youtu.be/C0eCJ-zp8K0

The build:
https://youtu.be/GJj2uyVsynI

I have a string that basically says do not write any code when we start a chat but write out your plan on accomplishing the task and ask me any clarifying questions.