LiteLLM is an open-source unified API that provides a single interface for calling 100+ LLM APIs including OpenAI, Anthropic, Gemini, Mistral, and local models, all in the OpenAI format. Developers can switch between providers with a single line change, implement fallbacks and load balancing, track costs across providers, and add rate limiting without changing their application logic. LiteLLM also provides a self-hosted proxy server for teams needing centralized API key management, budget controls, and access logging across their organization.