-no_install-7e8aa0?style=for-the-badge){kind=icon}
Most AI tools give you a chatbox. OwLLM gives you a workforce — composable teams of specialised agents that you build, own, and run on your hardware, your VPS, or in a VM. 24/7. Local models or cloud APIs, your choice. Privacy by default.
Local AI today is in roughly the place the web was in 1995 — the open-weight models are real, the consumer hardware is there, and the workstation hasn't been built yet. The local-LLM tools that exist (Ollama, LM Studio, Jan) are great chatboxes. None of them help you compose a team of agents, fine-tune a model, ship a custom GGUF, or run a 24/7 agentic service on a $5/month VPS. That's the gap OwLLM fills.
We believe:
- You should own your AI workforce, not rent it.
- The data layer is community-owned — agent teams, role definitions, model profiles, MCP server recommendations all live in public, fork-and-PR.
- Privacy isn't a feature, it's a default. The app is local-first. Nothing leaves your machine unless you explicitly send it.
- Power tools matter more than chat polish. Fine-tuning, abliteration, quantization, red-teaming — the things serious users actually need.
- Cloud models are peers, not bosses. Use them when they're the right tool; replace them when they're not.
| owllm | The flagship — desktop app, module registry, agent team templates, community contributions. Start here. |
More to come: headless / VPS runner, public team marketplace, vision module — see the roadmap.
- 💻 Use the app — latest release
- 💬 Talk — Discussions
- 🎨 Contribute — agent teams, roles, translations, docs via CONTRIBUTING.md
- 🐛 Report bugs — Issues
- ⭐ Star the repo — the single best signal that this category is worth investing further in
Your team of AI agents. Build them. Own them. Run them anywhere.