Multi Provider Token Counter - Statuslines

Features

• Automatic provider detection (Claude, GPT-4, Gemini, Grok)
• 2025 verified context limits (1M for latest models)
• Real-time usage percentage calculation
• Color-coded warnings (green <50%, yellow <80%, red ≥80%)
• Formatted token counts with thousand separators
• Provider-specific icons for visual identification
• Supports legacy models with correct limits
• Context window utilization trend tracking

Use Cases

• Monitoring context usage across multiple AI providers
• Tracking token consumption to avoid context limit errors
• Comparing model efficiency across providers
• Real-time awareness of remaining context budget
• Multi-model workflow optimization
• Cost optimization by selecting models with appropriate context limits

Requirements

• Claude Code CLI installed and configured
• Bash shell available (bash 4.0+ recommended for pattern matching with [[]] and string manipulation)
• jq command-line JSON processor (jq 1.6+ recommended for safe extraction with // defaults)
• awk command (for percentage calculation with floating point precision)
• bc calculator (optional, for precise floating point comparison - script has integer fallback)
• Terminal with ANSI color code support (256-color mode recommended for color-coded warnings)

Configuration

{
  "statusLine": {
    "type": "command",
    "command": "$CLAUDE_PROJECT_DIR/.claude/statuslines/multi-provider-token-counter.sh",
    "refreshInterval": 1000
  }
}

Examples

Enhanced Multi-Provider Token Counter with Input/Output Split

Extended version tracking input and output tokens separately

#!/usr/bin/env bash

# Enhanced Multi-Provider Token Counter with Input/Output Split

read -r input

model=$(echo "$input" | jq -r '.model.id // .model.display_name // "unknown"')
input_tokens=$(echo "$input" | jq -r '.cost.input_tokens // 0')
output_tokens=$(echo "$input" | jq -r '.cost.output_tokens // 0')
total_tokens=$((input_tokens + output_tokens))

# Determine provider and context limit (same logic as base script)
if [[ "$model" == *"claude"* ]] || [[ "$model" == *"Claude"* ]]; then
  if [[ "$model" == *"sonnet-4"* ]] || [[ "$model" == *"sonnet-3.5"* ]] || [[ "$model" == *"Sonnet 4"* ]]; then
    limit=1000000
    provider="Claude"
    icon="🔮"
  else
    limit=200000
    provider="Claude"
    icon="🔮"
  fi
elif [[ "$model" == *"gpt-4.1"* ]] || [[ "$model" == *"GPT-4.1"* ]]; then
  limit=1000000
  provider="GPT-4.1"
  icon="🤖"
elif [[ "$model" == *"gemini"* ]] || [[ "$model" == *"Gemini"* ]]; then
  if [[ "$model" == *"2."* ]] || [[ "$model" == *"1.5-pro"* ]]; then
    limit=1000000
    provider="Gemini"
    icon="💎"
  else
    limit=128000
    provider="Gemini"
    icon="💎"
  fi
elif [[ "$model" == *"grok-3"* ]] || [[ "$model" == *"Grok 3"* ]]; then
  limit=1000000
  provider="Grok"
  icon="⚡"
else
  limit=100000
  provider="Unknown"
  icon="❓"
fi

# Calculate usage percentage
tokens=$total_tokens
if command -v awk > /dev/null 2>&1; then
  percentage=$(awk "BEGIN {printf \"%.1f\", ($tokens / $limit) * 100}" 2>/dev/null || echo "0.0")
else
  percentage="0.0"
fi

# Color coding
if command -v bc > /dev/null 2>&1; then
  if (( $(echo "$percentage < 50" | bc -l 2>/dev/null) )); then
    color="\033[32m"
  elif (( $(echo "$percentage < 80" | bc -l 2>/dev/null) )); then
    color="\033[33m"
  else
    color="\033[31m"
  fi
else
  percentage_int=${percentage%.*}
  if [ $percentage_int -lt 50 ]; then
    color="\033[32m"
  elif [ $percentage_int -lt 80 ]; then
    color="\033[33m"
  else
    color="\033[31m"
  fi
fi

# Format token counts
if command -v printf > /dev/null 2>&1; then
  tokens_formatted=$(printf "%'d" $tokens 2>/dev/null || echo "$tokens")
  limit_formatted=$(printf "%'d" $limit 2>/dev/null || echo "$limit")
  input_formatted=$(printf "%'d" $input_tokens 2>/dev/null || echo "$input_tokens")
  output_formatted=$(printf "%'d" $output_tokens 2>/dev/null || echo "$output_tokens")
else
  tokens_formatted="$tokens"
  limit_formatted="$limit"
  input_formatted="$input_tokens"
  output_formatted="$output_tokens"
fi

RESET="\033[0m"

echo -e "${icon} ${provider} │ ${color}${tokens_formatted}${RESET}/${limit_formatted} (${percentage}%) │ In:${input_formatted} Out:${output_formatted}"

Multi-Provider Token Counter with Custom Limits

Version with configurable context limits via environment variables

#!/usr/bin/env bash

# Multi-Provider Token Counter with Custom Limits

read -r input

model=$(echo "$input" | jq -r '.model.id // .model.display_name // "unknown"')
tokens=$(echo "$input" | jq -r '.cost.total_lines_added // 0')

# Custom limits via environment variables (defaults to standard limits)
CLAUDE_LIMIT=${CLAUDE_TOKEN_LIMIT:-1000000}
GPT_LIMIT=${GPT_TOKEN_LIMIT:-1000000}
GEMINI_LIMIT=${GEMINI_TOKEN_LIMIT:-1000000}
GROK_LIMIT=${GROK_TOKEN_LIMIT:-1000000}
DEFAULT_LIMIT=${DEFAULT_TOKEN_LIMIT:-100000}

# Determine provider and context limit
if [[ "$model" == *"claude"* ]] || [[ "$model" == *"Claude"* ]]; then
  limit=$CLAUDE_LIMIT
  provider="Claude"
  icon="🔮"
elif [[ "$model" == *"gpt"* ]] || [[ "$model" == *"GPT"* ]]; then
  limit=$GPT_LIMIT
  provider="GPT"
  icon="🤖"
elif [[ "$model" == *"gemini"* ]] || [[ "$model" == *"Gemini"* ]]; then
  limit=$GEMINI_LIMIT
  provider="Gemini"
  icon="💎"
elif [[ "$model" == *"grok"* ]] || [[ "$model" == *"Grok"* ]]; then
  limit=$GROK_LIMIT
  provider="Grok"
  icon="⚡"
else
  limit=$DEFAULT_LIMIT
  provider="Unknown"
  icon="❓"
fi

# Calculate usage percentage
if command -v awk > /dev/null 2>&1; then
  percentage=$(awk "BEGIN {printf \"%.1f\", ($tokens / $limit) * 100}" 2>/dev/null || echo "0.0")
else
  percentage="0.0"
fi

# Color coding
if command -v bc > /dev/null 2>&1; then
  if (( $(echo "$percentage < 50" | bc -l 2>/dev/null) )); then
    color="\033[32m"
  elif (( $(echo "$percentage < 80" | bc -l 2>/dev/null) )); then
    color="\033[33m"
  else
    color="\033[31m"
  fi
else
  percentage_int=${percentage%.*}
  if [ $percentage_int -lt 50 ]; then
    color="\033[32m"
  elif [ $percentage_int -lt 80 ]; then
    color="\033[33m"
  else
    color="\033[31m"
  fi
fi

# Format token count
if command -v printf > /dev/null 2>&1; then
  tokens_formatted=$(printf "%'d" $tokens 2>/dev/null || echo "$tokens")
  limit_formatted=$(printf "%'d" $limit 2>/dev/null || echo "$limit")
else
  tokens_formatted="$tokens"
  limit_formatted="$limit"
fi

RESET="\033[0m"

echo -e "${icon} ${provider} │ ${color}${tokens_formatted}${RESET}/${limit_formatted} (${percentage}%)"

Multi-Provider Token Counter Installation Example

Complete setup script with provider detection testing and bc calculator verification

#!/bin/bash
# Installation script for Multi-Provider Token Counter

# Check for jq (required for JSON parsing)
if ! command -v jq &> /dev/null; then
    echo "Installing jq for JSON parsing..."
    if [[ "$OSTYPE" == "darwin"* ]]; then
        brew install jq
    elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
        sudo apt-get install -y jq || sudo yum install -y jq
    else
        echo "Please install jq manually: https://stedolan.github.io/jq/"
    fi
fi

# Check for bc (optional, for floating point comparison)
if ! command -v bc &> /dev/null; then
    echo "Installing bc calculator (optional, for precise percentage comparison)..."
    if [[ "$OSTYPE" == "darwin"* ]]; then
        brew install bc
    elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
        sudo apt-get install -y bc || sudo yum install -y bc
    else
        echo "bc is optional - script will use integer comparison fallback"
    fi
fi

# Test emoji/Unicode characters
if echo -e '🔮 🤖 💎 ⚡ ❓' &> /dev/null; then
    echo "Emoji characters supported: 🔮 🤖 💎 ⚡ ❓"
else
    echo "Warning: Emoji characters may not display correctly"
    echo "Install Nerd Font or emoji-capable font"
fi

# Test color support
if echo -e '\033[32mGreen\033[0m' &> /dev/null; then
    echo "ANSI color codes supported"
else
    echo "Warning: ANSI color codes may not work"
fi

# Test TERM variable
if [ -n "$TERM" ]; then
    echo "TERM variable set: $TERM"
else
    echo "Warning: TERM variable not set - colors may not work"
    echo "Set with: export TERM=xterm-256color"
fi

# Create statuslines directory
mkdir -p .claude/statuslines

cat > .claude/statuslines/multi-provider-token-counter.sh << 'SCRIPT_EOF'
#!/usr/bin/env bash

# Multi-Provider Token Counter - 2025 Context Limits
# Supports: Claude Sonnet 4 (1M), GPT-4.1 (1M), Gemini 2.x (1M), Grok 3 (1M)

read -r input

model=$(echo "$input" | jq -r '.model.id // .model.display_name // "unknown"')
tokens=$(echo "$input" | jq -r '.cost.total_lines_added // 0')

# Determine provider and context limit
if [[ "$model" == *"claude"* ]] || [[ "$model" == *"Claude"* ]]; then
  if [[ "$model" == *"sonnet-4"* ]] || [[ "$model" == *"sonnet-3.5"* ]] || [[ "$model" == *"Sonnet 4"* ]]; then
    limit=1000000
    provider="Claude"
    icon="🔮"
  else
    limit=200000
    provider="Claude"
    icon="🔮"
  fi
elif [[ "$model" == *"gpt-4.1"* ]] || [[ "$model" == *"GPT-4.1"* ]]; then
  limit=1000000
  provider="GPT-4.1"
  icon="🤖"
elif [[ "$model" == *"gemini"* ]] || [[ "$model" == *"Gemini"* ]]; then
  if [[ "$model" == *"2."* ]] || [[ "$model" == *"1.5-pro"* ]]; then
    limit=1000000
    provider="Gemini"
    icon="💎"
  else
    limit=128000
    provider="Gemini"
    icon="💎"
  fi
elif [[ "$model" == *"grok-3"* ]] || [[ "$model" == *"Grok 3"* ]]; then
  limit=1000000
  provider="Grok"
  icon="⚡"
else
  limit=100000
  provider="Unknown"
  icon="❓"
fi

# Calculate usage percentage
if command -v awk > /dev/null 2>&1; then
  percentage=$(awk "BEGIN {printf \"%.1f\", ($tokens / $limit) * 100}" 2>/dev/null || echo "0.0")
else
  percentage="0.0"
fi

# Color coding based on usage
if command -v bc > /dev/null 2>&1; then
  if (( $(echo "$percentage < 50" | bc -l 2>/dev/null) )); then
    color="\033[32m"  # Green
  elif (( $(echo "$percentage < 80" | bc -l 2>/dev/null) )); then
    color="\033[33m"  # Yellow
  else
    color="\033[31m"  # Red
  fi
else
  percentage_int=${percentage%.*}
  if [ $percentage_int -lt 50 ]; then
    color="\033[32m"
  elif [ $percentage_int -lt 80 ]; then
    color="\033[33m"
  else
    color="\033[31m"
  fi
fi

# Format token count with commas
if command -v printf > /dev/null 2>&1; then
  tokens_formatted=$(printf "%'d" $tokens 2>/dev/null || echo "$tokens")
  limit_formatted=$(printf "%'d" $limit 2>/dev/null || echo "$limit")
else
  tokens_formatted="$tokens"
  limit_formatted="$limit"
fi

RESET="\033[0m"

echo -e "${icon} ${provider} │ ${color}${tokens_formatted}${RESET}/${limit_formatted} (${percentage}%)"
SCRIPT_EOF

chmod +x .claude/statuslines/multi-provider-token-counter.sh

# Add to settings.json
if [ ! -f .claude/settings.json ]; then
    echo '{"statusLine":{"type":"command","command":"$CLAUDE_PROJECT_DIR/.claude/statuslines/multi-provider-token-counter.sh","refreshInterval":1000}}' > .claude/settings.json
else
    jq '.statusLine = {"type":"command","command":"$CLAUDE_PROJECT_DIR/.claude/statuslines/multi-provider-token-counter.sh","refreshInterval":1000}' .claude/settings.json > .claude/settings.json.tmp
    mv .claude/settings.json.tmp .claude/settings.json
fi

echo "Multi-Provider Token Counter installed successfully!"
echo "Note: Ensure terminal supports emoji characters for provider icons"
echo "Test with: echo -e '🔮 🤖 💎 ⚡ ❓'"

Troubleshooting

Percentage shows >100% or incorrect limit

Update model detection logic with latest model names. Check Claude docs for current model IDs. Verify model field: echo '$input' | jq .model.id. Test pattern matching: [[ "$model" == "claude" ]] && echo "Matched". Update limit values for new model versions. Check if tokens exceed limit (may indicate incorrect limit or token counting issue).

Icons showing as boxes or question marks

Install Nerd Font or emoji-capable font. Test with: echo '🔮 🤖 💎 ⚡ ❓'. Verify terminal encoding: locale charmap (should be UTF-8). Set encoding: export LANG=en_US.UTF-8. Check font configuration in terminal settings. If not supported, modify script to use ASCII alternatives: icon="[C]" for Claude, "[G]" for GPT, etc.

Colors not working or showing as escape codes

Ensure TERM environment variable supports colors: echo $TERM (should show 'xterm-256color' or similar). Set if needed: export TERM=xterm-256color. Test colors: echo -e '\033[32mGreen\033[0m' (should show green text). Check 256-color mode: tput colors (should be >= 256). Verify ANSI escape sequences are supported by terminal.

Model detection fails for new AI models or versions

Add patterns to detection logic. Common patterns: claude-_ (Opus/Sonnet/Haiku), gpt-_ (4o/4.1/o3), gemini-_ (2.0/2.5-pro), grok-_ (3/4). Update with: [["$model" == "pattern"]]. Check model name format: echo '$input' | jq .model.id. Test pattern matching: [["test-claude-sonnet-4" == "claude"]] && echo "Matched". Add new model variants to appropriate provider block.

Percentage calculation precision too low or rounded values

Increase awk precision for decimals. Current: awk 'BEGIN {printf "%.1f", ...}' (1 decimal). Use: awk 'BEGIN {printf "%.2f", ($tokens / $limit) _ 100}' for two decimals. Higher precision: awk 'BEGIN {printf "%.3f", ...}' (3 decimals). Note: Higher precision may slow execution. Verify calculation: echo '456789 1000000' | awk 'BEGIN {printf "%.1f", (456789 / 1000000) _ 100}' (should return 45.7).

Token count showing as 0 or incorrect values

Check JSON field names: echo '$input' | jq .cost.total_lines_added (tokens). Verify jq extraction: echo '$input' | jq -r '.cost.total_lines_added // 0'. Check if field exists: echo '$input' | jq 'has("cost")'. Verify token counting: Some Claude Code versions may use different field names (e.g., 'total_tokens', 'input_tokens', 'output_tokens'). Update script to check multiple field names.

bc command not found or floating point comparison fails

Install bc calculator: macOS (brew install bc), Linux (sudo apt-get install bc or sudo yum install bc). Verify installation: command -v bc (should return /usr/bin/bc or similar). Test bc: echo '45.7 < 50' | bc -l (should return 1 for true). If bc unavailable, script uses integer comparison fallback. Verify fallback works: percentage_int=${percentage%.*} (extracts integer part).

Thousands separator not displaying in token counts

Verify printf supports thousands separator: printf "%'d" 12345 (should return 12,345). Check locale settings: locale (should include thousands separator). Set locale if needed: export LC_NUMERIC=en_US.UTF-8. Test formatting: printf "%'d" 1234567 (should return 1,234,567). If not supported, script falls back to plain numbers. Alternative: Use awk for formatting: awk '{printf "%'"'"'d", $1}' <<< 12345.