Claude Code Statusline

A custom statusline script for Claude Code that shows real-time session information directly in your terminal — context window usage, session cost in USD and BRL, prompt cache efficiency, model name, and git branch.

Preview

[claude-sonnet-4-6] 📁 my-project | 🌿 main
████████░░ 78% of 200K | $0.42 (R$2,39) | ⏱️ 8m 15s
cache: 62K read (81% hit)

Three lines of information at a glance:

1. Model + working directory + git branch
2. Context window progress bar + percentage + window size + cost (USD and BRL) + session duration
3. Prompt cache hit rate and tokens read from cache

Features

• Context window usage bar (green → yellow → red as you approach the limit)
• Session cost in USD with live BRL conversion (fetched from AwesomeAPI, cached 24h, with offline fallback)
• Prompt cache hit rate — highlights when you're saving money via caching (90% cheaper than regular input tokens)
• Warning indicator when context exceeds 200K tokens
• Git branch display when inside a git repository
• Supports both 200K and 1M (extended) context windows

Installation

Quick install (one command)

curl -fsSL https://raw.githubusercontent.com/hugohvf/claude-code-statusline/main/install.sh | bash

This will:

1. Check that jq is installed
2. Download statusline.sh to ~/.claude/statusline.sh
3. Patch ~/.claude/settings.json to enable the statusline
4. Print confirmation — just restart Claude Code

Requires: jq — install with brew install jq (macOS) or sudo apt install jq (Linux)

---

Let Claude set it up for you

Already inside a Claude Code session? Paste this prompt and Claude will handle the entire setup:

Please install the statusline from https://github.com/hugohvf/claude-code-statusline for me.

Steps:
1. Download https://raw.githubusercontent.com/hugohvf/claude-code-statusline/main/statusline.sh to ~/.claude/statusline.sh and make it executable
2. Open ~/.claude/settings.json and add (or merge) this key:
   "statusLine": { "type": "command", "command": "~/.claude/statusline.sh" }
3. Confirm what you changed and tell me to restart Claude Code to activate it

---

Manual installation

1. Download the script

curl -o ~/.claude/statusline.sh https://raw.githubusercontent.com/hugohvf/claude-code-statusline/main/statusline.sh
chmod +x ~/.claude/statusline.sh

2. Configure Claude Code

Add the following to your Claude Code settings (~/.claude/settings.json):

{
  "statusLine": {
    "type": "command",
    "command": "~/.claude/statusline.sh"
  }
}

Tip: You can open settings.json quickly with /config inside Claude Code.

3. Restart Claude Code

The statusline will appear at the bottom of your terminal session.

Compatibility

This statusline is exclusive to Claude Code (the official Anthropic CLI). Other AI coding tools do not support custom statusline commands:

Tool	Statusline support
Claude Code	✅ Supported
Cursor	❌ Not supported
OpenAI Codex CLI	❌ Not supported
GitHub Copilot	❌ Not supported

---

Requirements

• bash — available by default on macOS and Linux
• jq — for JSON parsing. Install with brew install jq (macOS) or apt install jq (Linux)
• curl — for BRL exchange rate fetching (optional — falls back to a hardcoded rate if unavailable)
• git — for branch display (optional — silently skipped if not in a git repo)

How it works

Claude Code passes a JSON payload to the statusline command via stdin on every update. The script reads it using jq, extracts the relevant fields, formats them, and prints up to 3 lines to stdout.

The USD→BRL exchange rate is fetched once per day from AwesomeAPI and cached at /tmp/statusline-usd-brl. If the request fails or you're offline, it falls back to 5.70.

For more details on how the statusline feature works, see the official Claude Code statusline docs.

Customization

You can edit statusline.sh to:

• Change the BRL fallback rate (BRL_FALLBACK)
• Adjust the cache TTL (BRL_CACHE_MAX_AGE, default 24h)
• Change color thresholds for the context bar (currently yellow at 70%, red at 90%)
• Change cache hit thresholds (green at ≥70%, yellow at ≥30%, red below)
• Add or remove lines

License

MIT