Claude Code: Root Directory Tradeoffs | Generated by AI

Home 2026.05

Question: When using Claude Code launched from the root directory (/), can you use @ to reference files across all projects without copying paths? And will running it from / be slow?

Answer:

Yes, your intuition is correct — and there are real tradeoffs to understand.

How @ Works in Claude Code

When you use @ to reference a file in a prompt, Claude Code includes the file’s content directly into the conversation context. File paths can be relative or absolute, and you can use tab-completion after typing @ to quickly navigate and select files without typing full paths.

The @ reference syntax allows you to quickly add a file’s content to the context without Claude having to find and read the file itself, which usually takes longer than explicit referencing.

So yes — if you launch from /, you can @ any file on the system with tab-completion. You won’t need to copy paths from outside.

Will Running from / Be Slow?

Short answer: Yes, noticeably so. Here’s why:

1. @ Autocomplete Uses ripgrep to Scan the Directory

If @ file mentions aren’t finding files, the bundled ripgrep binary may be the culprit. Claude Code uses ripgrep to power file search and @ autocomplete. Starting from / means ripgrep has to index the entire filesystem every time you type @something, which will be very slow.

2. The File Picker Slowdown Is a Known Issue

If your project is a large monorepo where the default file picker is slow, you can replace it with a custom command via the fileSuggestion setting in settings.json:

{
  "fileSuggestion": {
    "type": "command",
    "command": "~/.claude/file-suggestion.sh"
  }
}

The command receives JSON via stdin with a query field and outputs newline-separated file paths.

3. @ Autocomplete Slowdown Is a Confirmed Bug Too

There is a known reported issue where entering files with @ used to be snappy but now takes nearly a second for matching files to appear — and that’s just in a normal project directory, not from /. From /, it would be far worse.

4. Context Window Fills Fast

The context window is the most important resource to manage. A single debugging session or codebase exploration might generate and consume tens of thousands of tokens. LLM performance degrades as context fills — Claude may start “forgetting” earlier instructions or making more mistakes.

Better Alternative: Use --add-dir

Instead of running from /, run Claude Code inside your primary project and add other project directories as needed:

Claude Code’s multi-directory support lets you work across multiple directories in one session using the --add-dir flag:

claude --add-dir /path/to/backend --add-dir /path/to/frontend

This lets you work across frontend, backend, and shared modules without switching sessions, while maintaining good performance.

Summary

Approach Cross-project @ Performance
Launch from / ✅ Can reference everything ❌ Very slow autocomplete, huge ripgrep scope
Launch from project + --add-dir ✅ Explicit multi-dir access ✅ Fast, scoped search
Launch from project dir only ❌ Limited to one project ✅ Fastest

Recommendation: Use --add-dir for cross-project work. It gives you the same convenience of referencing files from multiple projects without the performance penalty of scanning from /.

References:

Back Donate