Today we're launching claw.zip — a closed-loop token optimizer for the Claude API.
If you're building with Claude, you're probably spending more than you need to on API calls. Most prompts contain redundant tokens that don't affect output quality. And many queries get routed to expensive models when a cheaper one would produce identical results.
How it works
claw.zip sits between your application and the Claude API. When a query comes in, we:
- Compact — Strip redundant tokens while preserving semantic meaning
- Route — Score the query complexity and send it to the cheapest capable model
- Save — You pay only for what's needed
The result: 80–93% reduction in monthly spend for most users.
Why we built this
Token costs are the silent killer of AI projects. Teams prototype with Claude Opus, get great results, then panic when the invoice arrives. They switch to smaller models, quality drops, users complain. It's a lose-lose.
claw.zip breaks this cycle. You keep using the best model for each query — we just make sure you're not overpaying for it.
What makes it different
- Closed-loop optimization — We measure actual output quality, not just token count. If compression hurts results, we back off automatically.
- Smart routing — Not every question needs Opus. Simple queries go to Haiku, complex ones to Sonnet, and only the hardest problems hit Opus. You get the same quality at a fraction of the cost.
- Zero config — Drop in your API key, point your app at claw.zip, and you're saving money in under 60 seconds.
Get started
Sign up at claw.zip, grab your API key, and swap your base URL. That's it.
# Before
ANTHROPIC_API_URL=https://api.anthropic.com
# After
ANTHROPIC_API_URL=https://api.claw.zip
Your code stays the same. Your prompts stay the same. Your bill doesn't.