The Inference Diet Docs

Set up Clawzempic, understand the routing pipeline, and build smarter bots that cost less.

🔍 ⌘K
⚡ Quickstart 📡 API Reference 💻 CLI Commands 🧠 How Routing Works
Getting Started
Quickstart
Get running in under two minutes with a single command.
Get started →
What is Clawzempic?
A drop-in LLM proxy that cuts inference costs 70-95% with routing, caching, memory, and security.
Learn more →
How It Works
The optimization pipeline: scoring, routing, caching, windowing, memory, scripts.
Read more →
Features
Intelligent Routing
Scores complexity and routes to the cheapest model that can handle it.
Explore →
Prompt Caching
Automatic cache breakpoints save up to 90% on repeated context.
Explore →
God-Tier Memory
Persistent facts and preferences across sessions, installs, and channels.
Explore →
Context Windowing
Compresses long conversations so your bot never burns its budget.
Explore →
Model Cascades
Customizable 4-tier routing mix. Tune the IQ dial per-client.
Explore →
Security Shield
Injection detection, credential redaction, and tool inspection on every request.
Explore →
Reference
CLI Reference
init, test, status, savings, doctor, store-key, restore, flags.
View commands →
API Reference
Chat completions, models, pricing, insights, settings, and more.
View endpoints →
Integrations
OpenClaw, generic SDKs, environment variables, and provider setup.
View guides →
Help
Anthropic Provider
Direct Claude API with full prompt caching.
Read →
OpenRouter Provider
300+ models with Anthropic caching on compatible models.
Read →
Troubleshooting
Connection errors, key formats, common issues.
Get help →