• Skills
  • Projects
  • Changelog
  • Toolkit
SkillsProjectsChangelogToolkit

© 2026 Rahul Ragi.

My setup, my tools, and the stuff I actually use every day.

This is what my setup looks like right now. Hardware, software, AI tools. The stuff I use every day to write code and ship things.

Last updated February 2026

Workstation

  • MacBook Pro M3

    Switched to Apple Silicon a while back. The M3 runs everything I need, stays cool, and the battery lasts long enough that I stopped carrying a charger around.

  • Samsung Smart Monitor M7 (43")

    Honestly more of an entertainment screen than a workstation monitor. I bought it during COVID to watch shows and it stuck around. Not great as a monitor or a TV, but it works fine as a second screen since I do most of my work on the laptop anyway.

  • Apple Magic Keyboard

    The base model without Touch ID. The Touch ID version costs extra and I didn't feel like paying for it. Low-profile, rechargeable, and nice to have a proper keyboard when I'm at the desk.

  • Apple Magic Mouse

    I'm in love with the design but I'll be honest, I barely use it. The MacBook trackpad does everything I need, so the Magic Mouse mostly sits on the desk looking pretty. Still don't regret buying it though.

  • IKEA RODULF Standing Desk

    Got this to remind myself to stand up sometimes. The motor is quiet and switching between sitting and standing is quick enough that I actually do it.

  • IKEA LÅNGFJÄLL Chair

    Comfortable enough for long sessions and doesn't cost a fortune like most ergonomic chairs people recommend. The armrests help.

Development tools

  • Cursor

    Where I write code. It's basically VS Code so my extensions still work, but the AI stuff and MCP support on top make it way more useful.

  • Ghostty

    My terminal. Fast, looks good on macOS, and I didn't have to configure anything. The fonts render better than any other terminal I've tried.

  • Nushell

    Switched from zsh recently. Everything comes out as structured data instead of text blobs, so filtering and piping actually makes sense. The syntax is different but I'm getting used to it.

  • Google Chrome

    DevTools is still the best way to debug web stuff. I live in the network panel and performance tab.

AI assistants

  • Claude Code

    Runs in my terminal and does most of the heavy lifting. Scaffolding new features, refactoring stuff across files, writing commits. I hook it up to MCP servers so it can pull from Figma, docs, and run tests.

  • Claude Sonnet 4.5

    What I use for planning. I describe what I want to build and it helps me think through the approach before I start writing code.

  • Claude Opus 4.6

    For bigger changes that touch a lot of files. It gets things right more often than the smaller models when the task is complicated.

  • Gemini 3 Pro

    I ask it to review code or help debug when I want a second opinion. Different models notice different things.

  • Codex 5.3 Research

    OpenAI's research-focused coding model. Good for digging into unfamiliar codebases and exploring solutions I haven't considered.

MCP servers

  • Context7

    Feeds current docs into Cursor and Claude Code so they stop hallucinating old APIs. I just type "use context7" and it grabs whatever library docs I need.

  • Figma MCP

    Lets Claude see my Figma files directly. I point it at a frame and it writes components that actually match the design. Saves a lot of back and forth.

  • CodeRabbit

    Reviews my PRs automatically. It's caught real bugs before they made it to main, so I keep it running on everything now.

  • Playwright MCP

    I tell Claude what to test and it writes and runs Playwright scripts for me. It reads the page structure directly instead of looking at screenshots, which works better than I expected.

Productivity

  • Raycast

    Replaced Spotlight. I mostly use it for clipboard history and moving windows around, but it does a lot more that I haven't explored yet.

  • Rabbitholes

    Infinite canvas for AI conversations. Each node is its own chat with its own context, and you can connect them together. I use it to break problems into smaller pieces instead of one long thread. Pick whatever model you want per node. Data stays local.

  • Jotdo

    A todo app I made for myself. It's just a list focused on today. Overdue stuff gets pulled forward automatically, and there's a sidebar to look back at what I finished on other days.