Liter Llm

This library is a universal client for large language model APIs. It has a Rust core with native bindings for 14 languages, including Python, TypeScript, Go, Java, Ruby, PHP, C#, Elixir, Dart, Swift and Zig, and it reaches 174 providers such as OpenAI, Anthropic, Google, Bedrock, Azure and Mistral through one interface. Switching providers means changing the model name rather than rewriting code. The API covers chat and streaming, tool calling, embeddings and rerank, image generation, speech and transcription, moderation, search and OCR. The same binary also runs a proxy server that is compatible with the OpenAI API, so an existing OpenAI SDK can point at it and gain virtual keys, budgets and fallbacks. It also includes a built in MCP server for AI agents. It targets developers who want one integration across many model vendors. Routing can follow cost, latency or weight, with automatic fallback to another provider when one fails, set per request or once in the proxy. API keys are wrapped and redacted from logs, and cloud authentication for Azure, AWS and Vertex is handled with automatic token refresh. Installation is through package managers such as cargo, pip, npm, Maven, gem, composer and others, or as a Docker image and command line tool. The documentation site publishes no pricing and lists no paid plans.