Diode Nexus Overview
To get started, see Getting started.
Private AI models — organizationally aware AI
Diode Nexus (product brand: Protected AI) is how you plug self-hosted or partner-hosted Private AI into an organizationally aware package: chat, tools, channels, and policy — over Diode’s end-to-end encrypted network — not a public chatbot bolted onto the open internet. Nexus tunnels into your private model so it is never exposed to the public Internet.
The pitch is not “pick another SaaS LLM.” It is: bring the model your security team will approve, then ground it in your systems.
What you get
- LLMs and models — Configure any Private LLM (Ollama, private GPU stacks, partner-hosted endpoints). Built-in public support includes OpenAI, Anthropic, xAI, OpenRouter, and Perplexity. You choose how private and public models are routed.
- Organization-aware tooling — MCP and tools so answers come from your assets, registries, and APIs — not training cut-off alone.
- Protected agentic chat — Natural-language conversations with pipeline visibility while tools run.
- Meet users where they work — Diode Collab, web embed, ACP for apps (including Vibe), and an OpenAI-compatible API (alpha).
Why Private AI on Nexus
| Without Nexus | With Nexus |
|---|---|
| A private model endpoint sitting alone | Model + tools + channels + Diode path as one package |
| Org data never reaches the model safely | MCP / tools pull org context under your perimeter |
| Vibe / apps invent their own LLM wiring | ACP and workspace Nexus point at your Protected AI |
Configure providers under Settings → LLMs. See Private AI models and Configure an LLM.