After over a decade as a “terminal-based developer”, I ended up using Cursor for a while when we decided to make use of devcontainers at work. This series documents how I am “returning to my roots” using modern tools like DevPod, Zellij, and Claude Code AI assistance without leaving the command line.
NOTE: This is a living series. I’ll update posts as I discover new patterns, gotchas, and solutions. New posts may be added as the workflow evolves.
The Backstory
I’ve been a vim/terminal person ever since I started working with Ruby / Rails. I’ve also spent considerable time building tools to improve virtualized development environments - it all started when I went 100% on Vagrant in early 2013, transitioning to vagrant-lxc, building vagrant-cachier later that year, and ultimately building Devstep in 2014. Developer experience in virtualized workflows has been a consistent passion, but I stopped focusing on tooling when I shifted from consultancy to product companies.
When my current team adopted devcontainers, I ended up moving to Cursor because that’s what was easier and I wanted to experiment with its AI features. But the GUI felt painfully slow compared to my terminal workflow and I had to relearn a bunch of new shortcuts. Later, when I started using Claude Code, I ended up spending most of my time in Cursor’s built-in terminal.
I recently discovered DevPod and realized it solved my problem: SSH-based devcontainers without IDE lock-in. I could return to neovim, move to zellij (from tmux), and use my terminal tools while keeping the team’s devcontainer benefits. Claude Code, which I was already using in Cursor’s terminal, now works even better in a real terminal setup.
This is me scratching my own itch again - the latest evolution of a 10+ year journey improving my own virtualized dev environments, this time building on top of other people’s tooling instead of building my own :)
The Technical Stack
Here’s what I ended up with: DevPod gives me SSH access to devcontainers (instead of VSCode Remote Containers), running on standard Docker. The team keeps using Cursor/VSCode with .devcontainer/, I opt into .devcontainer-devpod/ - both read the same format, no disruption.
Inside the container I’m using Neovim (LazyVim config), Zellij for terminal multiplexing (tabs, panes, sessions), and Oh My Zsh for shell config. Claude Code runs via CLI for AI assistance, and I’m using git worktrees for parallel branch development. Everything’s mounted via Docker volumes so configs survive container recreation.
After a few weeks of using this setup, I’m sticking with it. The control and understanding are worth the friction for me. But it’s not necessarily better than Cursor/VSCode - it’s a different set of tradeoffs that happens to match my preferences.
If you’re happy with VSCode/Cursor, stay there - the IDE integration works well. But if you’re terminal-based and frustrated by devcontainer IDE requirements, read on. Part 1 walks through the DevPod setup.
The Series
The posts cover the setup, daily workflow, and tooling details:
Part 1: DevPod: SSH-Based Devcontainers Without IDE Lock-in
How to use DevPod to SSH into devcontainers instead of opening them in VSCode/Cursor. Covers installation, wrapper scripts, config persistence, git signing fixes, and team adoption using a dual-setup approach so you don’t disrupt teammates who prefer IDEs.
Part 2: Using Zellij and Claude Code Over SSH
What daily development actually feels like with Zellij (terminal multiplexer) and Claude Code CLI. Keyboard shortcuts, persistent sessions, the unlock-first workflow, and the honest friction points that IDE users don’t deal with.
Part 3: Neovim and LazyVim for Rails Development over SSH
Setting up Neovim with LazyVim for Rails development over SSH. Clipboard integration using OSC 52, Ruby LSP configuration with bundled gems (avoiding Mason’s global installation for Rails projects), multi-cursor editing, global find/replace, and performance considerations when using git worktrees.
Part 4: Working on Multiple Branches Without Losing My Mind
How I built a Git worktree manager for running multiple branches simultaneously - port allocation, database isolation, and branch-specific Claude Code context that survives container recreation.