OpenMake is a small team building open-source, local-first AI software for people who want to run models and tools on infrastructure they control.
OpenMake LLM is an MIT-licensed, self-hosted AI workspace for local and open-weight LLMs.
- Route OpenAI-compatible models through vLLM and LiteLLM.
- Run autonomous AI agents in persistent Docker sandboxes with checkpoints and approval boundaries.
- Connect built-in and external MCP (Model Context Protocol) tools.
- Use role-based models, multi-agent Discussion, cited Deep Research, and HTML/PDF/DOCX artifacts.
- Keep external models optional with bring-your-own-key (BYOK) providers.
Our development setup runs OpenMake and a loopback-only LiteLLM gateway on a Mac mini, with private vLLM inference on an NVIDIA DGX Spark GB10 over Tailscale. That is our real setup, not a hardware requirement: OpenMake works with compatible OpenAI-style endpoints.
OpenMake LLM currently requires Node.js 24, Docker, PostgreSQL, and an OpenAI-compatible model endpoint. Contributions, reproducible deployment reports, and feedback about local GPU and model combinations are welcome.
