Stop overpaying for LLM tokens. cheaprouter automatically routes your completion requests to the cheapest available provider across Anthropic, OpenAI, Google Gemini, Groq, Mistral, DeepSeek, and Alibaba Qwen — selecting in real time based on price, latency, and availability. BYOK — Bring Your Own Keys. Pass your provider API keys per request. cheaprouter holds no credentials and stores no content. The routing intelligence is the value. Five tools:

arbitrage_get_pricing — full price comparison table across all 7 providers for any tier and token volume arbitrage_estimate_cost — pre-flight routing decision before you commit a request arbitrage_route_completion — routes and executes your completion on the cheapest eligible provider arbitrage_provider_status — see which providers you have access to and their current model/pricing config arbitrage_get_history — cumulative spend and savings tracking across sessions

Three capability tiers map semantically equivalent models across providers — tier_fast, tier_balanced, tier_powerful — so you're always comparing apples to apples. Latency-aware routing. Pass latency_sensitive: true to automatically exclude high-latency providers (useful when Chinese providers add 300–400ms from your region). Free. Open source. No vendor lock-in.