harbor

Harbor is a CLI tool and companion app that manages self-hosted LLM infrastructure through Docker Compose, letting users spin up pre-configured stacks of AI services with short commands like `harbor up`. It orchestrates backends such as Ollama, llama.cpp, and vLLM alongside frontends like Open WebUI, and supporting services including SearXNG for web search and Speaches for text-to-speech and speech-to-text. The project connects to Claude through an MCP server integration, meaning Claude Desktop and Claude Code can interact directly with locally running models and tools managed by Harbor. With hundreds of services available across dozens of Docker Compose files, the project targets homelab enthusiasts, self-hosting practitioners, and developers who want to run and compare local LLMs without manually wiring together each component. A notable detail is that a single `harbor up searxng speaches` command adds web retrieval-augmented generation and audio capabilities to an already-running Open WebUI instance.

Lo que la gente pregunta sobre harbor