token-counter-cli

Token Counter

A simple, fast, and user-friendly command-line tool to count the number of tokens in a text file. It provides a progress bar for large files and displays the output in a clean, readable format.

This tool uses the tiktoken library, which is the same tokenizer used by OpenAI for its large language models.

Features

• Fast Tokenization: Leverages tiktoken for high-performance token counting.
• Progress Bar: A rich-powered progress bar shows the status of large files.
• Styled Output: Displays results in a clean, formatted table.
• Flexible Encoding Selection: Choose specific tiktoken encodings via a flag or an interactive menu.
• Multiple File/Directory Support: Count tokens across multiple specified files or all supported files within a directory.
• Exclusion Patterns: Exclude files or directories using glob patterns.
• File Extension Control: Override default file extensions or add new ones to customize which files are processed. Supports wildcard auto-discovery.
• Recursive Directory Scanning: Optionally scan subdirectories recursively for comprehensive project analysis.
• LLM Context Limit Comparison: Compare token counts against common Large Language Model context window limits. These limits are loaded from src/token_counter/llm_limits.json and can be customized.
• Stdin Support: Process text piped directly to the tool.
• Easy to Use: Simple command-line interface for quick use.

Installation & Usage

Option 1: Install from PyPI (Recommended)

pip install token-counter-cli

Option 2: Install from source

git clone https://github.com/puya/token-counter-cli.git
cd token-counter-cli
uv init
uv add typer rich tiktoken InquirerPy
uv pip install -e .

Usage

# Basic usage - count tokens using default 'cl100k_base' encoding
token-counter my_document.txt

# Count tokens from stdin
echo "Your text here" | token-counter

# Count tokens in multiple files
token-counter file1.txt file2.md

# Count tokens in all supported files within a directory
token-counter my_project_folder/

Usage Options

Model Selection

# Use a specific encoding model
token-counter my_document.txt --model p50k_base
token-counter my_document.txt -m p50k_base

# Interactively select the encoding from a list
token-counter my_document.txt --select-encoding
token-counter my_document.txt -s

File Extension Control

# Override default extensions (only process specified extensions)
token-counter my_project_folder/ --extension .xml,.yaml,.toml
token-counter my_project_folder/ -e .xml,.yaml,.toml

# Add new extensions to the default list
token-counter my_project_folder/ --add-extensions .log,.temp
token-counter my_project_folder/ -a .log,.temp

# Auto-discover all extensions in target directory
token-counter my_project_folder/ --add-extensions "*"
token-counter my_project_folder/ -a "*"

File/Directory Exclusion

# Exclude specific files or directories using glob patterns
token-counter my_project_folder/ --exclude "*.log" --exclude "node_modules/"
token-counter my_project_folder/ -x "*.log" -x "node_modules/"

Recursive Directory Scanning

# Scan directories recursively (includes subdirectories)
token-counter my_project_folder/ --recursive
token-counter my_project_folder/ -r

# Combine recursive with wildcard extension discovery
token-counter my_project_folder/ -r -a "*"

# Recursive scan with exclusions
token-counter my_project_folder/ -r -x "node_modules/" -x ".git/"

LLM Context Limit Comparison

# Compare token count against common LLM context window limits
token-counter my_long_article.txt --check-limits
token-counter my_long_article.txt -c

Combined Options

# Complex example combining multiple options
token-counter my_project_folder/ -s -x "*.test.py" -c -e .py,.js -r

# Process only Python files recursively, exclude tests, and check limits
token-counter . --extension .py --exclude "*test*" --check-limits --recursive

# Auto-discover extensions recursively with exclusions
token-counter my_project/ -r -a "*" -x "node_modules/" -x ".git/"

# Add log files to processing and use specific model
token-counter logs/ --add-extensions .log --model p50k_base --recursive

Complete Option Reference

Option	Short	Description
--model	-m	Specify the encoding model (e.g., 'cl100k_base', 'p50k_base')
--select-encoding	-s	Interactively select the encoding model from a list
--extension	-e	Override default file extensions (comma-separated)
--add-extensions	-a	Add to default file extensions (comma-separated). Use * for auto-discovery
--exclude	-x	Exclude files/directories using glob patterns (repeatable)
--check-limits	-c	Compare token count against LLM context window limits
--recursive	-r	Recursively scan subdirectories when processing directories
--help		Show help message and exit

Examples

```bash
token-counter test_article.txt
token-counter test_article.txt -m p50k_base
token-counter test_article.txt -s
echo "Hello world" | token-counter
token-counter README.md test_article.txt
token-counter src/
token-counter . --exclude "*.md" --exclude "src/"
token-counter test_article.txt -c
token-counter . --extension .py,.js
token-counter . -e .py,.js
token-counter . --add-extensions .log,.xml
token-counter . -a "*" 
token-counter . --recursive
token-counter . -r -a "*" -x ".git/"
```

Configuration Files

The token-counter tool uses two JSON configuration files to customize its behavior:

File Extensions Configuration (src/token_counter/config/allowed_extensions.json)

This file defines which file extensions are processed by default when scanning directories. The current default extensions include:

{
  "default_extensions": [
    ".txt", ".md", ".py", ".js", ".ts", 
    ".json", ".html", ".css"
  ]
}

Command-line overrides:

• --extension or -e: Override the default extensions entirely. Only files with the specified extensions will be processed.
• --add-extensions or -a: Add new extensions to the default list without removing the existing ones. Use * for auto-discovery.

If both flags are provided, --extension takes precedence and a warning will be displayed.

LLM Context Limits Configuration (src/token_counter/config/llm_limits.json)

This file contains context window limits for major LLM providers and models, used when the --check-limits flag is specified. The file includes the latest models from:

• OpenAI: GPT-4.1 series (1M tokens), GPT-4.5 (1M tokens), GPT-4o series
• Anthropic: Claude 4 Opus/Sonnet (200K tokens), Claude 3.7/3.5 Sonnet
• Google: Gemini 2.5 Pro/Flash (1M tokens), Gemini 1.5 Pro (2M tokens)
• Meta: Llama 4 Scout (10M tokens), Llama 4 Maverick (1M tokens), Llama 3.x series
• xAI: Grok 3 (~131K tokens)
• Mistral: Large 2, Medium 3, Small 3.1 (128K tokens)
• Cohere: Command A (256K tokens), Command R/R+ (128K tokens)

You can edit this file to add, remove, or modify models and their corresponding token limits to suit your needs.

Example usage:

token-counter large_document.txt --check-limits

This will show how your token count compares against all configured model limits.

Adding to PATH

To use the token-counter command from anywhere in your system, you need to add the virtual environment's bin directory to your shell's PATH.

• Get the full path to the bin directory:

  pwd
  # Copy the output and append /.venv/bin to it.
  # For example: /Users/you/token-counter/.venv/bin
  

• Add the path to your shell's configuration file:

  • For Bash (usually ~/.bashrc or ~/.bash_profile):

    echo 'export PATH="/path/to/your/project/.venv/bin:$PATH"' >> ~/.bashrc
    source ~/.bashrc
    

  • For Zsh (usually ~/.zshrc):

    echo 'export PATH="/path/to/your/project/.venv/bin:$PATH"' >> ~/.zshrc
    source ~/.zshrc
    

  Now you can run token-counter from any directory.