I'm really trying to figure out how to effectively use memory systems when coding. I've given a few methods a shot, like creating planning documents and trying to instruct Claude on my preferences and the artifacts needed for my projects. However, I feel quite frustrated because once I start coding, Claude tends to produce unexpected bugs and adds unnecessary features, making the initial plan quickly irrelevant. He often marks things as done too soon, even when they aren't finished or well-implemented. So, I end up having to rewrite the plans to reflect the actual state of things, which is just tedious.
I haven't tried the memory MCP tool because I'm struggling to find good examples specifically for coding. Most resources I've found focus on things like genealogy rather than practical coding applications, even though I can see how they might relate. My project is pretty big already, and trying to break it down into an entity relationship diagram is proving too complex (Gemini 2.5 barely managed it). Ultimately, I find my own memory of the project to be more effective than any memory tool I've tried. That said, I'm interested in knowing what memory systems or tools you all use and how they help you maintain structure in your projects across chats!
1 Answer
I use the MCP tool to coordinate tasks and keep everything organized in one place. It really helps manage the context of projects when working with a RAG agent, plus you can check it out here: [Agent-MCP](https://github.com/rinadelph/Agent-MCP).
Isn't that just another MCP tool that makes you use OpenAI for vectorizing data? Why not just use open-source models from Hugging Face instead?