AI Terminal

The AI tab gives you a chat panel with direct access to your server — you can ask questions, get command suggestions, and run approved commands without leaving KoreShell.

Opening AI

Click the AI tab on any connected server. The AI context is scoped to that server — the model knows the server hostname, OS, and any context you provide.

Configuring a Provider

Before sending your first message, configure an AI provider in Settings → AI Providers.

Supported Providers

Provider	Models	Notes
Anthropic	Claude Haiku, Sonnet, Opus	Default — recommended
OpenAI	GPT-4o, GPT-4o mini	Requires API key
Google Gemini	Gemini 1.5 Flash, Pro	Requires API key
OpenRouter	Any available model	Single key, many models
Ollama	Any local model	No API key — fully local

See AI Providers → for setup instructions per provider.

Setting an API Key

1. Settings → AI Providers → select provider → paste API key → SAVE
2. Keys are stored in the local vault (AES-256-GCM) — never transmitted except to the provider's API

Chatting

Type a message and press Enter (or Shift+Enter for a new line). Responses stream token by token.

Example prompts

• What's using all the disk space on this server?
• Show me the last 50 lines of /var/log/nginx/error.log
• Write a systemd service file for my Node app
• Why would CPU spike every 5 minutes?

Command Execution

When the AI suggests a command, KoreShell shows it in a consent block before running anything:

┌─────────────────────────────────────────┐
│ ▶ PROPOSED COMMAND                      │
│                                         │
│   df -h                                 │
│                                         │
│  [RUN]  [COPY]  [DISMISS]               │
└─────────────────────────────────────────┘

• RUN: executes on the server, streams output back to the chat
• COPY: copies to clipboard — run manually in the TERM tab
• DISMISS: skip this command

:::warning Commands need approval KoreShell never runs commands automatically. Every execution requires explicit RUN click. Destructive commands (rm, truncate, stop service) show an extra confirmation step. :::

Conversation History

Each server has its own conversation list. Previous conversations are saved in the local database — click any conversation in the sidebar to resume it.

The provider and model used for each conversation are preserved — replies always use the same provider the conversation was started with.

Switching Models Mid-Conversation

Use the model selector in the AI tab header to switch provider or model. The switch takes effect on your next message.

Ollama (Fully Local AI)

Ollama runs models entirely on your machine — no data leaves the device.

1. Install Ollama from ollama.ai
2. Pull a model: ollama pull llama3
3. In KoreShell: Settings → AI Providers → Ollama → set base URL (default http://localhost:11434)
4. Select Ollama as provider in the AI tab → pick a pulled model

Privacy

• AI conversations are stored locally in the KoreShell database
• Prompts and responses are sent to the provider's API over HTTPS — subject to their privacy policy
• Command output returned to the chat is not stored by KoreShell beyond the local conversation log
• No server credentials or vault data are ever included in prompts