AI Assistant

Termial's AI assistant helps you work more efficiently by generating commands, analyzing terminal output, and even executing multi-step tasks automatically.

AI Modes

Command Mode

The default mode for quick command generation:

1. Click the AI panel or focus it with a keyboard shortcut
2. Type what you want to do in natural language
3. Example: "find all files larger than 100MB"
4. Press Enter to generate the command
5. Review the suggestion and click to insert it into your terminal

Example prompts:

Context Mode (Pro)

Context mode includes your terminal output in the AI request, allowing the AI to understand what's happening:

1. Click the Context toggle to enable context mode
2. The AI will see recent terminal output
3. Ask questions like "why did this fail?" or "what does this error mean?"
4. Get contextual answers based on what's actually in your terminal

Use cases:

• Debugging errors
• Understanding command output
• Getting suggestions based on current state

Agent Mode (Pro)

Agent mode lets the AI execute multi-step tasks automatically:

1. Click the Agent button or switch to Agent mode
2. Describe a complex task in plain language
3. Example: "Install nginx and configure it as a reverse proxy for port 3000"
4. The AI creates an execution plan
5. Review and approve each step (or use auto-approval)
6. Watch as the AI executes the task

Agent capabilities:

• Execute shell commands
• Read and write files
• Navigate directories
• Upload and download files via SFTP
• Create port forwards/tunnels

Approval modes:

• Smart (recommended): Auto-approves safe commands, asks for dangerous ones
• Approve all: You approve every single step
• Auto (dangerous only): Auto-approves everything except destructive commands

Using the AI Assistant

Asking Questions

Type natural language questions or requests:

• "How do I find all Python files?"
• "Show me disk usage"
• "What's running on port 8080?"
• "Restart the nginx service"

Reviewing Suggestions

When the AI generates a command:

1. Review the command carefully
2. Click to insert it into your terminal
3. The command is inserted but not executed automatically
4. Press Enter to run it (or edit first)

Command History

Previous AI interactions are saved. Access them by scrolling up in the AI panel.

Agent Mode Details

Starting a Task

1. Open Agent mode
2. Enter your task description
3. Click Start Agent
4. The AI will probe your environment and create a plan

Execution Plan

The AI shows you its plan before execution:

• List of steps to complete
• What each step will do
• Estimated number of commands

Approving Actions

Based on your approval mode:

• Approve: Execute this action
• Reject: Skip this action
• Pause: Stop and wait
• Cancel: Stop the entire task

Task Continuation

After a task completes, you can continue the conversation:

• "Now also install SSL certificates"
• "Create a backup of the config first"
• The AI remembers context from the previous task

AI Settings

Configure AI behavior in Settings → AI Settings:

• AI Model: Choose which model to use
• API Configuration: Set up custom API endpoints if needed

Plan Limits

AI usage depends on your plan:

Need more? Purchase additional credits in the app.

Tips for Better Results

1. Be specific: "Find large log files in /var/log" works better than "find logs"
2. Provide context: "I'm on Ubuntu 22.04" helps get platform-specific commands
3. Use context mode: When debugging, enable context mode so AI can see your output
4. Start simple with Agent: Test with small tasks before complex multi-step operations