I've been diving into AI coding for a while, starting with tools like bolt.new and lovable, which made it super easy to throw together some prompts and get a front-end working smoothly. Lately, I've transitioned to using cursor and roo code. I really liked roo, but cursor just didn't work for me. If it wasn't for the high API costs, I'd still stick with roo since I can manage context and memory issues with some good planning. Now that I've switched to Claude Code, I'm curious about what tech stacks others are using to enhance the capabilities of CC. Personally, I'm utilizing Supabase CLI, MCP, and Fly.io for my backend, while I'm deploying my frontend on Netlify. With the recent updates of Opus and Sonnet 4, the tokens seem to deplete rapidly, so I'm looking for additional areas to optimize and keep making progress.
3 Answers
For front-end work, I usually start with v0 and then convert everything to HTML + Tailwind, using Claude Code to refine the design from that point. For data analysis and quick web scraping, I've set up a custom MCP server that lets Claude test and run one-off Python scripts. It's pretty efficient!
Today, I got a half-working tmux MCP server set up and successfully interacted with Claude Code from Claude Desktop. It’s been a fun project figuring things out!
I just got Claude Code up and running myself! I haven't completely figured out my workflow yet since I was mainly using Lovable before, but I'm trying to work on strategies like planning in Claude AI to avoid wasting tokens. It’d be great to hear from anyone else who isn’t too tech-savvy about how they’re adjusting their workflows!
I’ve been there! Sometimes it feels overwhelming, but just breaking it down into smaller tasks helps a lot. You might want to set up a basic structure and iterate from there.
That's interesting! How does MCP really fit into your workflow for data analysis? Can't you just run everything straight in Claude Code?