llm.rb is Ruby's capable AI runtime. It runs on Ruby's standard library by default. loads optional pieces only when needed, and offers a single runtime for providers, agents, tools, skills, MCP, A2A (Agent2Agent), RAG (vector stores & embeddings), streaming, files, and persisted state. As a bonus, llm.rb is also available for mruby. It supports OpenAI, OpenAI-compatible endpoints, Anthropic, Google Gemini, DeepSeek, xAI, Z.ai, AWS Bedrock, Ollama, and llama.cpp. It also includes built-in ActiveRecord and Sequel support, plus concurrent tool execution through threads, tasks (via async gem), fibers, ractors, and fork (via xchan.rb gem).