Simple pricing
Free forever for phrase compression. Advanced unlocks the Enigma machine.
TokenShrink Free
No credit card. No expiry.
- Web compressor
- Unlimited compressions
- API access + SDK
- All LLM providers supported
- Usage dashboard
- No sign-up required to try
- Rosetta Protocol (Enigma codec)
- Domain rotors (React / Node / Python / Supabase)
- Sub-agent vocab inheritance
- Cross-session learning
TokenShrink Advanced
$79/year — save 27%
- Everything in Free
- Rosetta Protocol (Enigma codec)
- Session codebook — negotiated cipher
- Domain rotors (React / Node / Python / Supabase)
- Sub-agent vocab inheritance
- Cross-session learning — Hall of Fame
- Compaction-safe Rosetta re-injection
- Advanced usage dashboard
What is the Rosetta Protocol?
The Enigma Machine concept: at session start, Claude and you negotiate a shared codebook — a compressed cipher built from your actual project's vocabulary. Every subsequent message is compressed using that cipher. The cost is paid once, amortized across every message in the session.
Domain Rotors
Auto-detects your stack from package.json. React, Node, Python, Supabase vocab packs load automatically — 88 terms across 4 domains.
Cross-Session Learning
Terms that fire frequently across sessions get promoted to the Hall of Fame. Dead weight gets pruned. The codec improves every session.
Compaction-Safe
Context window resets? The Rosetta block is re-injected automatically. Sub-agents inherit the session vocab. The cipher never drops.
Frequently asked questions
Is Free really free forever?
Yes. Phrase compression, API access, and the usage dashboard stay free with no credit card and no expiry. We mean it.
What does Advanced add on top of Free?
The Rosetta Protocol — a negotiated session cipher. At session start, a codebook is built from your project vocabulary. Every message is then compressed using that cipher. Advanced also includes domain rotors (tech-stack vocab packs), sub-agent vocab inheritance, and cross-session learning that improves the codec over time.
Does compression affect AI response quality?
No. We prepend a tiny decoder header that teaches the LLM our abbreviations. The AI understands the compressed prompt just as well as the original. For prompts under 30 words, we skip compression entirely.
What AI providers does it work with?
All of them. OpenAI, Anthropic, Google, Mistral, Llama, Cohere — any LLM that accepts text prompts. Our SDK has first-class support for OpenAI and Anthropic.
Do you store my prompts?
No. We never store your prompt text. We only store word counts, compression ratios, and usage statistics. Your prompts are processed in memory and immediately discarded.
Can I cancel Advanced at any time?
Yes. Cancel from the billing portal in your dashboard. You keep Advanced features until the end of your billing period, then drop back to Free automatically.