Connecting a model
The model toolbar at the top of the Agent view controls which LLM answers your prompts. Switch providers or models anytime; new messages use the current selection.
Toolbar overview
- Provider dropdown — Ollama, OpenAI, Anthropic, Gemini, Groq, and others you configure
- Model dropdown — models available for the selected provider
- Token / cost hint — session usage for the active conversation
- Settings gear — opens Settings → AI Providers
A Local only pill appears when cloud calls are blocked by policy.
Select a provider and model
- Ensure at least one provider is configured (see Ollama setup or Cloud providers).
- Open the Agent view.
- In the toolbar, choose your provider (e.g. Ollama).
- Choose a model from the second dropdown.
If the model list is empty, open Settings, verify the provider, and click Refresh models (Ollama) or Verify (cloud).
Recommended defaults for solo use
| Goal | Suggestion |
|---|---|
| Free, private, offline-capable | Ollama + a pulled 7B–8B tool-capable model |
| Stronger reasoning / Simulink edits | Cloud model with native tool support (e.g. GPT-4 class, Claude) |
| Mixed | Ollama for exploration; switch to cloud for complex model_edit tasks |
Model switch and context
When you change models mid-session, MBD Mind can:
- Replay prior turns to the new model (default), or
- Summarize context via a handoff summary (Settings → Agent → handoff mode)
Long threads may hit context limits. Start a new agent session (Ctrl+N) for unrelated work.
Provider status indicators
In Settings → AI Providers, each provider shows a status:
- verified — credentials work; models listed
- not configured — add API key or start Ollama
- auth failed — recheck key and base URL
Budget awareness
Cloud providers show estimated spend against your optional monthly budget in AI Providers settings. Ollama reports tokens but $0 cost.
Next steps
- Reading usage — monthly breakdown by provider
- MATLAB MCP overview — model tools independent of which LLM you pick