TokTab CLI

LLM pricing data at your fingertips

A command-line interface for accessing TokTab, a free API providing pricing data for 2000+ LLM models. Powered by LiteLLM and updated nightly.

Installation

# Using uvx (recommended)
uvx toktab gpt-4o

# Or install with pip
pip install toktab

Usage

Get pricing info for a specific model

toktab gpt-4o
toktab claude-3-opus
toktab gemini-1-5-flash

Output:

╭──────────────────────────────────────────────────────────────────────────────╮
│ gpt-4o (openai)                                                               │
╰──────────────────────────────────────────────────────────────────────────────╯

Pricing
 Type    Cost / 1M tokens 
 Input              $2.50 
 Output            $10.00 

Context Window
 Limit       Tokens 
 Max input      128K 
 Max output      16K 

Capabilities
✓ Vision · ✓ Functions · ✓ Tool choice · ✓ System msgs

Search for models

toktab search claude
toktab search "gemini 3"
toktab search provider:anthropic

JSON output

All commands support --json for machine-readable output:

toktab --json gpt-4o
toktab search --json claude

Options

Options:
  --json     Output raw JSON
  --version  Show version
  --help     Show this message and exit.

AI Agent Integration

TokTab CLI is designed to work well with AI agents and automated pipelines.

Auto-detection

When stdout is not a TTY (e.g. piped to another process), output switches to JSON automatically — no flags needed.

# Human at a terminal gets rich tables
toktab gpt-4o

# Piped to another process gets JSON
toktab gpt-4o | jq .input_cost_per_token

Environment variable

Set OUTPUT_FORMAT to force a specific format:

export OUTPUT_FORMAT=json  # Always JSON
export OUTPUT_FORMAT=text  # Always rich tables

Schema introspection

Agents can discover CLI capabilities at runtime:

toktab schema

This outputs a machine-readable JSON description of all commands, arguments, options, and API endpoints.

Structured errors

When JSON output is active, errors are written to stderr as structured JSON:

{"error": true, "message": "Model 'nonexistent' not found"}

Model Slugs

Model identifiers are derived from LiteLLM model names with special characters replaced by hyphens.
For example:

• gemini/gemini-pro → gemini-gemini-pro
• anthropic/claude-3-opus → anthropic-claude-3-opus

Use the search command to find the exact slug for a model.

Features

• 🚀 Fast: Lightweight CLI with minimal dependencies
• 📊 Rich output: Beautiful tables with cost color-coding (green=cheap, yellow=medium, red=expensive)
• 🔍 Fuzzy search: Find models by name or provider
• 💰 Cost per million tokens: Easy-to-read pricing format
• 🎨 JSON output: Perfect for scripting and automation
• 🆓 Free: No API key required

Development

# Clone the repo
gh repo clone tomdyson/toktab-cli
cd toktab-cli

# Install with dev dependencies
uv venv
source .venv/bin/activate
uv pip install -e ".[dev]"

# Run tests
pytest

# Test locally
toktab gpt-4o

Publishing a New Release

1. Update the version in pyproject.toml:

   version = "0.2.0"  # Bump version number
   

2. Commit and push your changes:

   git add pyproject.toml
   git commit -m "Bump version to 0.2.0"
   git push
   

3. Create and push a git tag:

   git tag v0.2.0
   git push origin v0.2.0
   

4. Create a GitHub release:

   gh release create v0.2.0 --title "v0.2.0 - Release Title" --notes "Release notes here"
   

5. Done! GitHub Actions will automatically build and publish to PyPI.

The package will be live at pypi.org/project/toktab within 1-2 minutes.

License

MIT License - see LICENSE for details.

Credits

• Data sourced from LiteLLM
• API provided by TokTab