Why a Terminal Beats an IDE for Claude Code — My Daily TreeRU Workflow

Something struck me the very first day I used Claude Code. The AI writes great code — but everything around it lags behind.

An IDE like Cursor solves some of this. But IDEs are heavy. Switching projects means closing and reopening windows, and remote server work needs yet another layer of configuration. I wanted everything handled inside the terminal — lightweight and fast. So I built TreeRU, and the workflow I use today came together around it and Claude Code.

You only need three things.

git clone https://github.com/treeru/treeru.git
cd treeru && npm install
node index.js

Press Ctrl+Shift+D in Windows Terminal and the pane splits in half. Run Claude Code on the left, run treeruon the right. That’s the base layout.

┌─────────────────────────┬────────────────────────────────┐
│                         │ > ..         │ > components    │
│  Claude Code            │ > src        │   index.js      │
│                         │ > public     │   package.json  │
│  "edit this file"       │ > lib        │   README.md     │
│  "look at this error"   │   .env       │   tsconfig.json │
│  "run the tests"        ├──────────────┤                 │
│                         │              │                 │
│                         │  TreeRU      │                 │
└─────────────────────────┴──────────────┴─────────────────┘

The AI writes code on the left while you watch the file tree live on the right. TreeRU detects file changes automatically, so anything Claude Code creates or modifies shows up on the right immediately. No more typing ls or tree over and over.

This is the feature I reach for most. Register a project folder once, and from anywhere two keystrokes launch Claude Code inside that project.

One-time registration

The daily launch (F12)

Opening the file explorer, hunting for the folder, copying the path, opening a terminal, running cd, then typing claude — all of it collapses into F12 → Enter. It pays off most when you juggle several projects. I keep 5–6 registered — the blog, TreeRU, server admin, and so on — and switch between them constantly. Workspaces you no longer use can be deleted right from the F12 menu with the Del key.

Do you know the most frustrating moment when working with Claude Code? When an error shows up in the browser and you want to show the AI.

In an IDE you would just drag the image in, but in a terminal you have to hand over a file path as text. That was tedious for me at first too: capture, open the file explorer, save the image, copy the path… Since building TreeRU, here is how it goes.

The whole thing takes under 5 seconds. No browser, no file explorer, no typing out paths. Capture → copy path → paste. Three steps, and that’s it. It would barely be an exaggeration to say I built TreeRU for this one feature.

There’s one thing that nags at you while using Claude Code: “how much usage do I have left?” On a Pro or Max plan there are per-session and weekly limits, so you find yourself wanting to check the remaining budget constantly.

Normally that means opening a browser and going to claude.ai → Settings → Usage. It breaks your flow. So I built F8 into TreeRU.

  TreeRU v1073                          Session 12% 2:53 | Weekly 5%  3 SSH hosts
┌──────────────────────────────────────────────────────────────────────────────┐

Press F8 and two pieces of information appear in the top header:

• Session 12% 2:53 — current session usage is 12%, with 2 hours 53 minutes until reset
• Weekly 5% — weekly all-models usage is 5%

Color tells you the status at a glance too: below 50% is blue, 50–80% is yellow, above 80% is red. Red is your cue to start rationing.

One-time setup

The first time you press F8, a Chrome (or Edge) window opens. Log in to claude.ai. Close the browser. Press F8 again. Done.

The login session is stored locally, so from then on F8 is all you need. That first run takes about 3 seconds, but every refresh after that takes 0.4 seconds, because a headless browser stays alive in the background. Your login survives a reboot — only the first F8 is slightly slow, and it’s fast from there.

Sometimes you need to run Claude Code directly on a server — editing deploy scripts, changing server config, or chasing a bug that only reproduces in the remote environment.

TreeRU makes that dead simple.

Three steps — ssh user@server, cd /path/to/project, claude — collapse into a single F12 → Enter. If you have several servers, register each one. Local and remote projects share the same F12 list, so switching is fast.

Take a screenshot while connected over SSH and it uploads to the remote server automatically. Copy a local file with Ctrl+C, press F5, and it transfers to the remote folder over SFTP. No separate FTP client needed.

Here is my actual routine, unedited.

Honestly, there are clearly situations where an IDE is better. Autocomplete, debugger integration, Git GUIs — those are real strengths. But with Claude Code in the picture, the calculus changes.

The point is this: because Claude Code takes over the job of writing code, IDE autocomplete and IntelliSense matter comparatively less. What matters more is the supporting work alongside the AI — browsing files, handing over screenshots, switching projects— and that is exactly where the terminal + TreeRU combination is efficient. Above all, it’s light. Nothing heavy churning in the background.

None of this is an argument that IDEs are bad. I just wanted to make the case that if Claude Code is your primary tool, the terminal is more than enough — more than you might expect.

TreeRU is not an ambitious tool. File browsing, automatic screenshot saving, usage checks, SSH connections. Each piece is small, but combined with Claude Code they add up to an environment where you never miss having an IDE.

I built it myself and use it daily — this very blog post was written with this workflow. It’s open source and free, so give it a try with no strings attached.