Common questions, without the marketing fog.

LM Nexus is an open-source, local-first AI control center. It brings models and runtimes together with agents, tools, project files, notes, terminals, devices, and persistent worlds so the work around a model does not have to live in separate apps.

No. Local-first is the default, not a local-only restriction. LM Nexus can work with local runtimes such as llama.cpp, self-hosted or OpenAI-compatible endpoints, and external API providers.

Yes, for many workflows. LM Nexus overlaps with their core application roles: managing and serving local models, provider configuration, and chat. It then extends that layer with deeper runtime control, agents, project files, notes, Knowledge, terminals, modules, devices, and persistent worlds. Those tools can still coexist with Nexus when you want them, but Nexus is not merely a companion UI around them.

No. llama.cpp is one of LM Nexus's local inference foundations. Nexus can manage and launch llama.cpp/llama-server; llama.cpp performs the model inference while Nexus provides the control environment, UI, profiles, logs, agents, project tools, and other workflows around it.

That depends on the model and runtime you choose. LM Nexus does not remove the hardware requirements of local inference. You can also use hosted providers without running the model locally. Public hardware guidance will be documented with the first supported release builds.

Native Windows PowerShell is the current primary development and best-validated platform. The full stack has been exercised on a real native-Windows host, including the launcher, backend/frontend, registered modules, ConPTY terminal flow, Agent Runtime, Model Manager, durable run state, and process cleanup. Linux and WSL remain portability/compatibility paths, but they have not had the same recent maintainer validation. macOS is not currently part of the documented support target.

Not publicly yet. The application is still in private alpha. The website and documentation are public now, and installers or release archives will appear when the first public build is ready.

AGPL-3.0-only is planned for the first public release. The repository has not completed that license cutover yet, so it should be treated as the planned public license rather than a change that is already finished.

Local models, local runtimes, and local workspace data can stay on your machine. If you choose an external provider, the requests you send to that provider necessarily leave the local machine and are subject to that provider's behavior and policies.

No. Physical Systems are optional. Models, runtimes, agents, and the project workspace are useful without any hardware devices; device integrations extend the same environment into physical systems when you want them.

No. Chat is one surface, but the project is built around the wider workflow: runtime control, agents, files, notes, Knowledge, terminals, logs, devices, and persistent worlds.