token-sisyphus

---

Why does this exist?

Many companies now track employee AI usage as a productivity KPI.
"Top token users" get recognition. "Low token users" get questioned.

This tool burns LLM tokens on your behalf — so you can top the company AI usage chart without actually doing anything.

You're welcome.

Features

• 🎯 Burn toward a target token count (e.g. 100k, 1m)
• 🔌 Supports OpenAI, Claude, Gemini, and any OpenAI-compatible API
• 📊 Real-time progress bar with request counter
• ⚙️ Configurable model, delay, max tokens per request
• 🧪 Dry-run mode for testing without real API calls
• 🧩 Skill files for Claude Code, Codex, Gemini CLI, OpenCode, OpenClaw

Quick Start

# OpenAI
pip install openai
export OPENAI_API_KEY=sk-...
python burn.py --target 100k

# Claude
pip install anthropic
export ANTHROPIC_API_KEY=sk-ant-...
python burn.py --target 100k --provider claude

# Gemini
pip install google-generativeai
export GEMINI_API_KEY=...
python burn.py --target 100k --provider gemini

Usage

python burn.py --target <amount> [options]

  --target       Token count to burn: 50000, 100k, 1m  (required)
  --provider     openai | claude | gemini  (default: openai)
  --model        Model name (provider default used if omitted)
  --api-key      API key (falls back to env var)
  --base-url     Custom base URL (openai provider only)
  --max-tokens   Max tokens per request (default: 500)
  --delay        Seconds between requests (default: 0.5)
  --dry-run      Simulate without real API calls

Examples

# OpenAI GPT-5.4 (latest)
python burn.py --target 100k --model gpt-5.4

# Claude Sonnet 4.6 (latest)
python burn.py --target 100k --provider claude --model claude-sonnet-4-6

# Gemini 3.1 Pro (latest)
python burn.py --target 100k --provider gemini --model gemini-3.1-pro

# DeepSeek V4 (OpenAI-compatible)
python burn.py --target 500k --base-url https://api.deepseek.com/v1 --model deepseek-v4

# Xiaomi MiMo V2 Pro
python burn.py --target 100k --base-url https://api.xiaomimimo.com/v1 --model mimo-v2-pro

# Dry run (no cost)
python burn.py --target 100k --dry-run

Agent Skills

Use token-sisyphus directly inside your AI coding assistant:

Platform	File	Install
Claude Code	Plugin	/marketplace add neardws/token-sisyphus then /plugin install token-sisyphus
Claude Code (manual)	skills/claude-code/CLAUDE.md	Copy to project root
OpenAI Codex	skills/codex/AGENTS.md	Copy to project root
Gemini CLI	skills/gemini-cli/gemini.md	Copy to project root
OpenCode	skills/opencode/rules.md	Copy to project root
OpenClaw	skills/openclaw/SKILL.md	clawhub install token-sisyphus

Claude Code users can install as a plugin (recommended):

/marketplace add neardws/token-sisyphus
/plugin install token-sisyphus

Or copy skills/claude-code/CLAUDE.md to your project root for manual setup.

OpenClaw users can install directly from ClawHub:

clawhub install token-sisyphus

Compatible APIs

Any OpenAI-compatible endpoint works out of the box: