Click, run, work.
Download or select a model, click Run, and start working through a clean UI without terminal-first setup.
CONNECT • CONTROL • ORCHESTRATE
LM Nexus
Open-source local-first AI control center
Click-and-run simplicity for local models. Raw runtime control for power users. A modular workspace for agents, files, notes, knowledge, terminals, and add-ins.
30-second explanation
One local-first workspace for the layers around AI work.
Local AI workflows are scattered across model runners, chat UIs, terminals, notes, files, agents, scripts, and provider dashboards. LM Nexus brings these layers into one local-first modular workspace.
Local-first does not mean local-only. LM Nexus is designed to connect local runtimes, self-hosted servers, custom endpoints, and API providers from one control center.
Three modes
Simple when you want it. Inspectable when you need it.
Expose the runtime.
Inspect runtime settings, launch raw inference servers, view logs, copy endpoints, and connect directly when you need full control.
Bring the work together.
Use models inside a workspace with agents, files, notes, knowledge, terminals, logs, and modular workflows.
CONNECT • CONTROL • ORCHESTRATE
The product structure follows the brand phrase.
Models and providers
Connect local models, self-hosted runtimes, custom endpoints, and external/API providers.
Runtimes and behavior
Control model settings, runtime profiles, servers, raw endpoints, provider configuration, logs, and workspace behavior.
Workspace workflows
Orchestrate agents, files, notes, knowledge, terminal workflows, and modular add-ins from one workspace.
Runtime control
Raw runtime visibility without giving up an approachable UI.
LM Nexus is built around a model and provider layer with a built-in model manager, Hugging Face LLM/model browser, raw llama.cpp server orchestration, OpenAI-compatible provider support, custom endpoints, and logs/runtime visibility.
Direct raw-server access with no additional Nexus inference layer required. When users connect directly to a llama.cpp server started by LM Nexus, inference traffic does not pass through a Nexus inference proxy. In that mode, Nexus acts as an orchestrator/manager rather than an inference middleware layer.
curl http://127.0.0.1:PORT/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "local-or-custom-model",
"messages": [
{ "role": "user", "content": "Explain this workspace." }
]
}'- Built-in model manager
- Hugging Face LLM/model browser
- Raw llama.cpp server orchestration
- Direct raw endpoint visibility
- OpenAI-compatible provider support
- Custom endpoints
- Logs and runtime visibility
- Nexus API path for integrated workflows
Modular workspace
Models belong near the work they support.
LM Nexus includes chat, agents, chat delegations, files, notes, notes canvas, notes graph, knowledge, terminal, logs, modules, and future add-ins as parts of one modular AI workbench.
Chat and agents
Run model conversations, configure agents, and delegate work across chat contexts.
Files and knowledge
Use files, knowledge workflows, and provider context without leaving the workspace.
Notes and graphs
Capture notes, map them on a canvas, and explore 2D or 3D graph views.
Terminal and logs
Keep runtime status, logs, terminals, and settings visible for power-user workflows.
Private alpha screenshots
A first look at the workbench.
Screenshots are copied from the current LM Nexus app assets. Public builds are planned after the private alpha stabilizes.
Comparison preview
LM Nexus is designed to sit between runners, UIs, runtimes, and workspaces.
| Tool | Best at | LM Nexus difference |
|---|---|---|
| Ollama | Simple local model serving | Keeps click-and-run simplicity while adding deeper runtime visibility, raw server orchestration, profiles, logs, provider configuration, and workspace integration |
| LM Studio | Polished local model UX | Open-source, modular, and oriented around direct runtime/server orchestration |
| Open WebUI | Self-hosted AI interface | Goes lower by managing/exposing local runtimes and higher by connecting models to agents, notes, files, knowledge, terminals, and add-ins |
| llama.cpp | Raw inference runtime | Makes low-level control approachable with UI, profiles, logs, endpoints, and workflows |
| Agent tools | Automation/coding workflows | Provides the local-first workspace and runtime/provider layer around agents |
Roadmap preview
Private alpha first. Public builds after stabilization.
Now private alpha
- private alpha stabilization
- model manager
- runtime/server orchestration
- direct raw endpoint access
- providers
- files/notes/knowledge workflows
- agents integration
- install/onboarding polish
Next public alpha
- public alpha
- add-in architecture
- capability/action registry
- runtime profiles
- better logs/observability
- cross-module workflows
- docs and examples
- UI polish
Later future
- marketplace
- team workflows
- sync/backup
- enterprise deployment
- advanced automation
- optional cloud services
Support
Open-source project support will come with the public alpha.
LM Nexus is planned as an open-source project. Support options such as GitHub Sponsors may be added as the public alpha matures.