I've been working with Roo code in my cursor setup for about two months now, and overall, it has worked well for me. Currently, I'm utilizing the Roo memory bank along with custom flow agents, which has significantly lowered my API costs. However, I'm mainly using Sonnet 4 at the moment and I've been considering a switch to Claude Code since I'm spending around $40 a day on API calls. I'm curious, how does Claude Code handle context for projects? Is there a specific MCP you're using to assist with that? Also, any tips for someone thinking about making the switch would be greatly appreciated!
1 Answer
I've really been enjoying Claude Code. I've moved from Roo and found it simpler; it works great within cursor. It's just a more reliable tool for me, especially with tool calls and applying diffs. I don’t use memory actively, but it has memory built-in. I regularly maintain the claud.md files as necessary, and I’ve managed to reduce my need for multiple MCP servers—now, I’m just using the Playwright one. It handles built-in fetch and git tools without any issues!
Nice! Does it manage context well? I'm really fond of the Roo code memory bank because I don't stress over the 200K context limit. If things start to get convoluted, I create a new task, and it seamlessly accesses the memory bank to continue where it left off, without needing me to re-explain everything.