Thanks for reporting a bug you have found in Cursor!
Please add the following info to help us diagnose your issue:
Check the forum to ensure the issue hasn’t been reported already Provide a clear description of the bug Explain how to reproduce the bug (if known) Attach screenshots or recordings (e.g., .jpg, .png, .mp4). Tell us your operating system and your Cursor version (e.g., Windows, 0.x.x). Tell us if the issue stops you from using Cursor.
Hi all. I’m new in this forum, but I’ve been facing a problem. When I’m using Cursor, I see this screen many times:
Hi @deanrie thanks for the answer. Just one question: If I rename my project and opening it I won’t have the previous chat, is that correct? But in case I need the previous chat, (not the whole chat, but the most recent, ie: today’s chat) how can I do to delete older chats?
Most users also run custom instructions to build a memory bank or documentation, which the agent can then reference as needed instead of relying on context in chats. Which do not use the full context window of the model.
Personally, I delete chats multiple times a day and only use a single chat at any given time. I’m also on macOS.
I used SpecStory for a bit, but with the way I have rules and prompts set up, my agent keeps track of its own memory bank and documentation. Each time I start a new session, I use the summary feature, and with the prompts in place, I simply need to prompt it with:
The primary rules are in Cursor Settings (very top right gear) > Rules > User Rules. Anything you put there will be sent with every message you send.
You also have Cursor Rules, which you create. Cursor makes setting it up more manageable in Cursor Settings > Rules > Project Rules.
When you add a rule there, it will create a folder with your project’s root names.
.cursor/rules/
Any new rule file will be put there.
Read up on it here: (Use the AI Search to find answers; it works very well.)
There are plenty of systems here that work very well. Check out RIPPER-5.
I’m in the process of building a bot on Poe.com that will help with this. Once complete, it will have access to my collection of prompt engineering white papers. I’ll share it here once it’s finished.
I’d recommend loading your rules and all of the prompts for the memory bank in either Claude Projects, Poe.com, with context management disabled, or Open WebUI with context set to max so the LLM can see everything. ChatGPT Projects doesn’t load everything into context; it uses RAG instead. If you use Open WebUI, just be sure that if you attach text files, you enable Using Entire Document.
Then, simply tell it you want to incorporate the Cline memory bank into your platform, give it the documentation for Cursor rules and User Rules, and then have it handle revamping everything.
I also create my own agents instead of relying on the ones crafted by Cursor. Not that theirs aren’t great; I simply prefer having more control (). I created a multi-mode agent that uses inspiration from the RIPPER-5 prompt that @robotlovehuman released, which is excellent on its own. I also created a documentation agent. I’m pretty much set on almost everything I do between those two. The multi-mode agent has seven primary modes and two sub-modes, which are used with a few of the primary modes, giving a total of 14 modes that it can swap between on its own, less the documentation agent, as Cursor cannot currently swap between agent modes, at least as far as I know.
The sub-modes help enforce autonomous mode when I can’t babysit it or interactive mode when I want to actively participate in it or when it might need my intervention.