DEV.co
MCP Servers · av

harbor

Harbor is a CLI tool that orchestrates a complete local LLM stack—including Ollama, llama.cpp, vLLM frontends, Open WebUI, and supporting services like SearXNG and ComfyUI—via Docker Compose. A single `harbor up` command pre-wires these services so they work together without manual configuration.

Source: GitHub — github.com/av/harbor
3.1k
GitHub stars
211
Forks
Python
Primary language
Apache-2.0
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repositoryav/harbor
Ownerav
Primary languagePython
LicenseApache-2.0 — OSI-approved
Stars3.1k
Forks211
Open issues56
Latest releasev0.5.2 (2026-06-14)
Last updated2026-06-21
Sourcehttps://github.com/av/harbor

What harbor is

Harbor manages Docker Compose orchestration, environment configuration, and inter-service networking for local LLM deployments. It supports multiple backends (Ollama, llama.cpp, vLLM, DMR, MLX), frontends (Open WebUI), and specialized services (web search, voice chat, image generation, coding agents). Recent versions add agentic modules and workflow presets.

Quickstart

Get the harbor source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/av/harbor.gitcd harbor# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Local LLM development and experimentation

Developers and ML engineers needing a quick, pre-configured multi-service LLM stack for local prototyping without cloud dependencies or manual Docker setup.

Homelab and self-hosted AI infrastructure

Non-commercial homelabs and internal teams wanting a reproducible, one-command LLM stack with web search, voice, and image generation integrated locally.

Offline/air-gapped AI workflows

Organizations requiring completely offline LLM inference with integrated RAG, voice I/O, and image generation—no external API calls or cloud connectivity.

Implementation considerations

  • Requires Docker and Docker Compose installed; verify Docker daemon is running and user has appropriate permissions.
  • Disk space varies by backend and models chosen; llama.cpp defaults consume less than Ollama. Plan for 10–100+ GB depending on model sizes.
  • Initial `harbor up` pulls Docker images and may take several minutes on first run; subsequent starts are faster.
  • Service inter-connectivity is pre-wired, but custom model loading, fine-tuning, or token/resource limits require editing YAML or `.env` files.
  • Monitor port conflicts (Open WebUI defaults to 3000–3001, Ollama to 11434, SearXNG to 8888); use `harbor doctor` to detect and resolve.

When to avoid it — and what to weigh

  • Production SaaS deployments — Harbor is designed for local development and homelabs, not for cloud-scale production infrastructure or multi-tenant services.
  • Minimal resource environments — The full stack (multiple LLM backends, Open WebUI, SearXNG, ComfyUI, etc.) requires significant disk, memory, and CPU; unsuitable for lightweight deployments.
  • Users unfamiliar with Docker or Linux CLI — Despite the simplicity of `harbor up`, troubleshooting, custom service configuration, and `.env` management assume Docker and shell command familiarity.
  • Proprietary or closed-source model integration — Harbor focuses on open-source and self-hosted models; integration with commercial API-based models (OpenAI, Anthropic) requires additional wrapper layers.

License & commercial use

Apache License 2.0 (Apache-2.0): permissive OSI license allowing commercial use, modification, and distribution with attribution and liability disclaimer.

Apache-2.0 permits commercial use. However, individual bundled services (Ollama, Open WebUI, SearXNG, etc.) carry their own licenses (MIT, AGPL, GPL variants); review each service's license if incorporating into a commercial product. Harbor itself does not restrict commercial deployment.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityClear
Deployment complexityModerate
DEV.co fitGood
Assessment confidenceHigh
Security considerations

Local-only by default (services bind to localhost/127.0.0.1 unless explicitly exposed). No built-in authentication for Open WebUI in default config; enable if exposing to network. Models are downloaded from public registries (Ollama, Hugging Face); verify provenance. Docker container isolation provides baseline process separation; host filesystem access and GPU passthrough introduce standard container security considerations. No formal security audit or vulnerability disclosure policy documented.

Alternatives to consider

Ollama + manual Docker Compose

Offers more control and minimalism; suitable if you only need Ollama backend and are comfortable writing Docker Compose YAML yourself.

LLaMA.cpp + Gradio / Streamlit UI

Lighter weight, fewer dependencies, and more portable for single-model inference; trade orchestration features for simplicity.

LM Studio (GUI desktop app)

Cross-platform GUI alternative for local inference (macOS, Windows, Linux); no CLI/Docker required, but less suitable for headless or remote access scenarios.

Software development agency

Build on harbor with DEV.co software developers

Install Harbor and spin up a fully integrated LLM environment in seconds. Explore the wiki, join Discord for community support, and start experimenting with local inference today.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

harbor FAQ

Can I run Harbor on macOS or Windows?
Harbor CLI runs on macOS and Windows (via WSL2) if Docker Desktop is installed. Docker Compose orchestration works the same, though GPU acceleration varies by platform.
Do I need to download models manually?
No; Ollama, llama.cpp, and other backends auto-download models on first use. Specify model names in config or via CLI; Harbor handles the rest.
How do I add a custom service or model?
Edit the generated `docker-compose.yml` and `.env` files in your Harbor workspace, then restart with `harbor up`. Requires familiarity with Docker Compose syntax.
Is Harbor suitable for production?
Not in its current form. Harbor is designed for local development, homelabs, and experimentation. Production deployments require custom Kubernetes, SLA setup, monitoring, and hardening.

Software development & web development with DEV.co

Need help beyond evaluating harbor? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and mcp servers integrations — and maintain them long-term.

Ready to simplify your local LLM stack?

Install Harbor and spin up a fully integrated LLM environment in seconds. Explore the wiki, join Discord for community support, and start experimenting with local inference today.