ATC - Anthropic Tokens Counter

A command-line tool to count tokens for Anthropic's Claude models before sending requests.

Features

• Count tokens for various Claude models (Opus, Sonnet, Haiku)
• Read input from files or stdin
• Model management with short and full names
• Configurable default model
• Cross-platform XDG configuration support

Installation

From Source

git clone <repository-url>
cd claude-count-tokens
pip install .

For Development

git clone <repository-url>
cd claude-count-tokens
pip install -e .

Setup

Before using ATC, you need to set your Anthropic API key:

export ANTHROPIC_API_KEY="your-api-key-here"

Usage

Count Tokens

Count tokens from stdin:

echo "hello world" | atc

Count tokens from a file:

atc CLAUDE.md
atc ./CLAUDE.md
atc ~/.claude/CLAUDE.md
atc /Users/foo/.claude/CLAUDE.md

Count tokens with a specific model:

echo "hello world" | atc --model sonnet
atc CLAUDE.md -m haiku

Model Management

List available models:

atc models

List models with full names:

atc models --verbose
atc models -v

Set default model:

atc models set-default sonnet

Available Models

Short Names

• opus - Claude Opus 4.1 (default)
• opus-4, opus-4.1 - Claude Opus 4.1
• sonnet - Claude Sonnet 4.1
• sonnet-4, sonnet-4.1 - Claude Sonnet 4.1
• sonnet-3.7 - Claude Sonnet 3.7
• haiku - Claude Haiku 3.5
• haiku-3.5 - Claude Haiku 3.5
• haiku-3 - Claude Haiku 3

Full Names

• claude-opus-4-1-20250805
• claude-sonnet-4-1-20250805
• claude-sonnet-3-7-20240805
• claude-haiku-3-5-20241022
• claude-haiku-3-20240307

Configuration

ATC stores its configuration in $XDG_STATE_HOME/atc/settings.json (or ~/.local/state/atc/settings.json if XDG_STATE_HOME is not set).

The configuration file stores:

• Default model selection

Requirements

• Python 3.8+
• anthropic Python library
• Valid Anthropic API key

License

MIT License