llm-router

LLM Router is a local desktop application that acts as an intelligent proxy between AI clients (such as Claude Code, Cursor, and Codex CLI) and various LLM providers including OpenAI, Anthropic, Google Gemini, and DeepSeek. Its primary goal is to reduce API costs by maintaining conversation "stickiness" to a single provider, ensuring that prompt caching discounts are preserved during long conversations. The app provides a unified local API endpoint (localhost:11411) and supports both OpenAI and Anthropic style request formats. Key capabilities include: - Cost Optimization: Tracks tokens used, money spent, and savings achieved through cache hits. - Intelligent Routing: Customizable rules to route requests based on model names, prompt length, or provider availability. - Transparency: A "Routing Explain" feature that details why a specific model was chosen and what alternatives were considered. - Easy Integration: One-click "App Takeover" for specific tools like Claude Code and Codex CLI to automate configuration. - Privacy: Local-first architecture using an embedded SQLite database; API keys are stored locally and never leave the device. - Cross-Platform Support: Available for macOS, Windows, and Linux.