av/harbor

Stop configuring your AI stack. Start using it. One command brings a complete pre-wired LLM stack with hundreds of services to explore.

What it solves

Setting up a local AI environment is complex, requiring manual configuration of multiple backends, frontends, and supporting services. Harbor automates this orchestration, providing a pre-wired ecosystem of tools that work together out of the box.

How it works

Harbor uses a CLI and a companion application to manage Docker Compose orchestration. Users can use commands like harbor up to launch specific services (like Ollama or vLLM) or entire stacks. It handles configuration, cross-service connectivity, and even provides host-native inference for macOS (via MLX/oMLX) to bypass container overhead. It also includes a "Boost" module to chain agentic workflows and a tunneling service to securely expose local services to the internet.

Who it’s for

  • Developers wanting to run local LLMs and coding agents (like Codex or Aider) without manual config.
  • Users looking for a complete, private AI stack including web search (SearXNG), image generation (ComfyUI), and voice chat (Speaches).
  • Engineers needing to orchestrate complex agentic workflows and multi-step research tasks.

Highlights

  • One-command setup: Launch backends, frontends, and satellites with simple CLI commands.
  • Extensive service catalog: Supports major engines like vLLM, llama.cpp, and Ollama, plus UIs like Open WebUI and LibreChat.
  • Agentic workflows: Includes "Boost" modules for automated web research, audits, and scope checks.
  • Flexible connectivity: Provides OpenAI-compatible backends for host-side coding tools and MCP ecosystem support.