Letronna is an AI agent that runs locally on macOS, Windows, and Linux while also offering a hosted cloud environment for persistent tasks and web app deployment. It supports any OpenAI-compatible endpoint, Anthropic models, Ollama, llama.cpp, and a hosted model pool, switching between them mid-session. The agent can execute real commands, browse the web, manage files, and deploy live web apps at a public URL under letronna.si. A built-in wallet pays for API costs automatically, and plain-text memory is stored on disk between sessions. The core app and agent are MIT-licensed, with paid tiers adding hosted credits and cloud compute.