Cloud AI
Getting Started
NeuroFlash & Neuro
Reference
Cloud AI (OpenRouter)
Optional cloud backend alongside the built-in local llama engine. Connect to 300+ models with a single API key — disabled by default.
Why Optional Cloud
NeuroTerm ships with a local llama engine that runs entirely on your machine. For most commands and diagnostics, local is fast, private, and free. When you want a bigger model for a harder problem — deep multi-step debugging, complex reasoning, long-context analysis — OpenRouter gives you access to 300+ cloud models through one provider.
The cloud backend is disabled by default. You explicitly opt in per install, and can disable it again at any time.
Setup
Get an OpenRouter API key
Sign up at openrouter.ai and create an API key.
Open Settings → AI → Cloud Backend
Paste the key. NeuroTerm stores it encrypted (AES-256-GCM, with a key derived from this machine's ID) in its own credential file; on Windows it is also saved in Credential Manager. Configuration exports never include it.
Pick a default model
Choose from frontier models like GPT-4o, Claude Sonnet, Gemini Pro, Qwen, and hundreds more. You can switch models per request.
Use it anywhere AI is available
Neuro Input, Neuro Tools, and Auto-Explain all respect the backend choice. A visible indicator shows when cloud AI is active.
Privacy Model
Here is what leaves your machine while cloud AI is on, and what stays local. Settings → Advanced → Data & Privacy in the app lists every destination it sends to and whether it is on now.
- API key is stored encrypted on this computer and sent only to OpenRouter
- RAG embeddings are computed and stored locally
- Pattern markers and filters run locally
- Prompts, terminal text, connection names (host or serial port) and imported documents go to your chosen provider via OpenRouter, with secrets redacted
- Idle summaries, the session journal and error explanations send automatically while cloud AI is on
- A visible indicator shows when cloud AI is active
When to Use Cloud vs Local
| Use case | Recommended backend |
|---|---|
| Offline or air-gapped environments | Local only |
| Quick command lookup | Local (faster, free) |
| Privacy-sensitive codebase | Local |
| Deep multi-step debugging | Cloud (bigger models) |
| Long-context datasheet analysis | Cloud (larger context window) |
| Routine boot log review | Local |
Supported Providers
OpenRouter aggregates 300+ models from every major provider. Popular choices include:
- OpenAI — GPT-4o, GPT-4o-mini, o-series reasoning models
- Anthropic — Claude Sonnet, Claude Haiku, Claude Opus
- Google — Gemini Pro, Gemini Flash
- Meta — Llama 3 and Llama 4
- Alibaba — Qwen 2.5 and Qwen 3
- Plus Mistral, DeepSeek, Cohere, and many more
See the full list at openrouter.ai/models.