selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

Best self-hosted AI

Self-hosted AI software runs models, assistants, and inference engines on hardware you own or control, rather than on a vendor's servers. More about AI

Picking a model to run on your own GPU? See which open-weight LLMs fit your hardware →

--sort
374 projects· by score

** Today's picks

new every day · not on page 1

Projects

30 of 374

About self-hosted AI software

Ollama runs LLMs locally on macOS, Windows, and Linux and exposes a REST API with official SDKs. llama.cpp is a C/C++ inference engine that runs quantized models on CPU and GPU with minimal dependencies. vLLM focuses on throughput, using PagedAttention for LLM inference. Open-WebUI connects to Ollama and OpenAI-compatible APIs and adds RAG, plugins, and enterprise auth on your hardware.

378 self-hosted AI projects are listed here, ranked by popularity, maintenance and documentation.

Highest rated
Ollama (187 202 GitHub stars, MIT license, written in Go), followed by openclaw and llama.cpp
Docker
11 of the top 12 ship an official Docker image
Easiest to set up
Ollama and OmniRoute
Recently updated
openclaw v2026.11.4 on 2026-11-16

Top self-hosted AI software compared

ProjectLicenseLanguageStarsDockerSetupLatest releaseScore
OllamaMITGo187 202YesEasyv0.40.1694
openclawMITTypescript394 569YesModeratev2026.11.493
llama.cppMITC++135 258YesModeratev1.0.093
ComfyUIGPL-3.0Python147 141—Moderatev0.45.092
OmniVoice-StudioAGPL-3.0Python111 668YesModeratev0.5.891
firecrawlAGPL-3.0Typescript202 228YesAdvancedv2.11.091
Open-WebUIBSD-3-ClauseSvelte158 812YesModeratev0.13.390
vLLMApache-2.0Python97 025YesModeratev0.33.0rc390
RAGFlowApache-2.0Go94 239YesAdvancedv1.0.089
OmniRouteMITTypescript87 951YesEasyv3.8.5389
paperclipMITTypescript114 258YesModeratev2026.1106.088
AstrBotAGPL-3.0Python47 033YesModeratev4.30.0-alpha.188