AI providers & features
Local-first AI — Ollama, LM Studio, AntBot — plus optional cloud providers, Explain Screen, and the privacy monitor.
Providers
Local (default). Ollama, LM Studio, and AntBot work without any cloud API. Models are auto-detected from the running service, and AntBot's gateway can auto-start with xNAUT. Local-provider HTTP is routed through the Rust backend, so the webview's CORS restrictions never get in the way — the Test button in Settings probes the endpoint and reports Connected/Failed.
Cloud (optional). Anthropic, OpenAI, OpenRouter, and Perplexity — bring your own API keys, stored locally. Agent Chat can enumerate models from any configured OpenAI-compatible endpoint and groups them by provider in the model selector; each request routes through the provider that owns the selected model.
Your provider/model choice in the right-pane Agent Chat is persisted in xNAUT's settings and survives agent switches, navigation, pane remounts, and restarts.
AI features
- Explain Screen — reads the visible terminal output and explains what's happening; ideal for unfamiliar error walls.
- AI Theme Generator — describe a vibe ("submarine at dawn"), get a full ANSI theme for the entire app.
- Snippet Explain — one click explains any saved command.
- Plan Mode & Librarian — document-writing agents for planning and vault management (see Vault).
- Excalidraw MCP — xNAUT can clone, build, and run the official MIT-licensed Excalidraw MCP server locally on loopback; Agent Chat discovers its drawing tools automatically. A hosted endpoint + bearer key are optional overrides.
Privacy Monitor (ClawProxy)
A transparent LLM API proxy that inspects outgoing prompts before they leave the app:
- Detects leaked API keys, credentials, and PII
- Real-time privacy indicator in the status bar
- Privacy assessment panel with per-alert details
Combined with local providers, the default posture is: nothing sensitive leaves your machine, and if something would, you hear about it first.