Okay so ever since I found out about Cursor - my life changed. I was hooked. Common coding challenges became a thing of the past, but just like with everything else in life.. we always tend to find something wrong.
As I’m sure a lot of you experience it constantly - Cursor’s agent composer seems to be in the middle of doing something or moving to a new task and it completely forgot what it was last doing. This can obviously be a huge problem when you’re trying to create something that works together as a system. So I went off the deep end trying to find a solution.
My initial approach was trying to use Cursor’s rules to create somewhat of a scratchpad system to jot down important milestones and always look at them.. but it didn’t really work so well. It would rarely jot things down and even more rarely look at them.
I then went on to trying to create a javascript implementation to connect to the internal javascript engine Cursor runs on and implement hooks throughout runtime conversations to retrieve and store memory.. created crazy amounts of integrations and inputted a ton of time but in the end it didn’t run/integrate with Cursor how I thought it would. I tried this implementation over and over with no luck.
Then came MCP servers and the new .mdc rules… Finally I felt like I was on to something.. And my feeling was right 
After trying multiple different implementations - both local db’s and remote - I came across an awesome solution that would allow for my multi-dimensional memory system to come to life (awesome cause it was free for this use case and the API worked seamlessly). Turso using it’s API/SDK, libsql, and it’s vector embeddings.
I created an MCP server with different tools for different pieces of conversation and material, from storing user input and retrieving context at the beginning of every conversation to actively tracking files being created/edited, to storing episodes, milestones, requirements, etc all the way through the end of the conversation storing all the agent/assistant messages etc.
Now to make it autonomous, I created extremely precise and strict Cursor rules to attach to the system so it would ALWAYS use the tools throughout every conversation.
Long story short… I have a working solution that I would love to get feedback and improve upon with the community. I call it Cursor10x.
Along with the creation of the MCP server, I created a project with the .mdc rules and the mcp.json file implemented. All you have to do is download or copy and paste the .cursor/rules directory, the .cursor/mcp.json file, then create a free database using Turso and plug in your db url and auth token credentials into the mcp.json file.
I got excited/creative and baked in a task-management workflow/system into the rules. I then decided to create a website for people to be able to simplify their workflow when starting a brand new project… cursor10x.com where you can use my chatbot (uses Gemini 2.5 Pro in the backend) to enter details about your project - it then creates a technical blueprint, file/folder architecture, implementation guide and full list of detailed tasks with prompts (to use with the task management system) and lets you download it all as a .zip file with the Cursor10x system already included.
This whole project is free and open-source.. would love to get feedback from the community as I said so we can continue improving it and making Cursor even better than it already is Check out the links below!
I’ll make a post once the new project is deployed! It will be called DevLaunch. Also making some major improvements to cursor10x MCP but now it will be called DevContext (it applies to more than just cursor now). Thanks for the feedback.