polyrouter

Polyrouter offers an open-source, self-hosted gateway for managing various large language models (LLMs) from a single endpoint. Key capabilities include: * Centralized model routing via a single URL and API key * Explicit-first routing with configurable fallbacks * Hard spend limits to control expenditures * Transparent, metadata-only cost tracking * Self-hosting for complete data control and zero markup This platform allows developers to point their applications and agents to one unified interface, simplifying model integration and management across different providers like OpenAI, Anthropic, Groq, and local Ollama instances. It ensures reliability by honoring explicit model requests while providing clever routing options such as tier headers and "auto" selection with structural, semantic, and cascade layers for optimal model choice and graceful degradation. Polyrouter is designed for engineering teams and developers who require a robust, flexible, and cost-aware system for interacting with diverse models. It's ideal for those building agentic workflows, internal tools, or applications where precise control over model usage, spend, and data privacy is paramount. The dashboard provides clear insights into requests, spend by model, success rates, and provider health, helping teams manage their model consumption effectively.