A CLI application that uses AI to process natural language commands and execute them in the terminal.
It supports openAI gpt-4o, gpt-4o-mini and llama3.2 through ollama. It will support claude and gemini soon.
npm install -g terminal-ai-tool
If you are using ollama, make sure the server is running and the llama3.2 model is downloaded.
Run the initialization command to configure your AI provider:
ai setup
This will guide you through setting up your preferred AI provider and API key. The configuration will be stored in ~/.terminal-ai/config.yaml.
In non-interactive mode, the AI converts your natural language request into a single terminal command and executes it:
ai "your command in natural language"
ai "list all files in the current directory"
ai "delete all files in current folder"
When using a potentially destructive command, the application will ask for confirmation before executing.
In interactive mode, the AI maintains a continuous conversation, suggesting and executing commands with your permission:
ai --interactive "your task description"
Or use the shorthand:
ai -i "your task description"
The agent will:
- Analyze your request and suggest appropriate commands
- Provide reasoning for each suggested command
- Ask for confirmation before executing potentially dangerous commands
- Return command output to the AI for further analysis
- Maintain conversation context for multi-step tasks
To exit agent mode, type exit or quit.
ai -i "help me find large files on my system"
ai -i "set up a basic Node.js project"
The AI CLI can process input from other commands via pipes:
# Process file contents
cat config.json | ai "validate this JSON file"
# Filter command output
ps aux | ai "find all Node.js processes"
# Analyze logs
tail -n 100 server.log | ai "what errors occurred in the last 100 lines"Threads allow you to maintain conversational context across multiple interactions with the AI. This is especially useful for complex tasks that require multiple steps or when you want to continue a conversation later.
Each thread maintains its conversation history, allowing the AI to reference previous commands and responses. This provides continuity and context awareness for ongoing tasks.
To view all available threads:
ai thread listYou can filter threads by name:
ai thread list --filter "project"After listing threads, you can select one interactively to continue the conversation.
To continue a conversation in an existing thread:
ai thread attach <thread-id>This will resume the conversation where you left off, with all previous context intact.
To use a specific thread for a one-time command without entering agent mode:
ai --thread <thread-id> "your command"To give a thread a more descriptive name:
ai thread rename <thread-id> "React project setup"When you're done with a thread, you can delete it:
ai thread delete <thread-id>-
Start a new task in agent mode:
ai -a "set up a React project"This creates a new thread automatically.
-
Later, list your threads to find the React project:
ai thread list
-
Attach to the thread to continue the setup:
ai thread attach <thread-id>
-
After completing the task, rename the thread for future reference:
ai thread rename <thread-id> "React project setup"
Profiles allow you to manage multiple AI provider configurations, making it easy to switch between different providers or models without reconfiguring your settings each time.
Each profile contains:
- A unique name
- An AI provider (OpenAI, Claude, Gemini, or Ollama)
- An API key
- A selected model
The active profile determines which AI provider and model is used for processing your commands.
To view all configured profiles:
ai profile listThis shows all profiles with their provider, model, and pricing information. The active profile is marked.
To switch to a different profile:
ai profile set <profile-name>All subsequent commands will use this profile until you change it again.
To remove a profile you no longer need:
ai profile delete <profile-name>If you delete the active profile, another profile will automatically be set as active.
To see all available models with their pricing information:
ai profile modelsYou can filter models by provider:
ai profile models --provider openaiYou can temporarily use a different profile for a single command:
ai --profile <profile-name> "your command"Or use the shorthand:
ai -p <profile-name> "your command"-
Run the setup wizard to create your first profile:
ai setup
-
Create additional profiles for different providers:
# The setup command will guide you through creating a new profile ai setup -
Switch between profiles as needed:
ai profile set work-openai ai profile set personal-claude
-
Use a specific profile without switching:
ai -p work-openai "analyze this log file"
Cost Tracking is not completly correct but it will give you an idea of the cost of your commands.