The CLIver CLI supports two modes: one-shot queries and interactive chat. Both go through the same AgentCore engine.

One-Shot Mode

Pass a prompt directly:

cliver "What time is it in Tokyo?"
cliver -p "Summarize README.md" -f README.md

Interactive Mode

Start a conversation session:

cliver

In interactive mode, you get:

• Multi-turn conversation with context maintained across turns
• Rich TUI with syntax highlighting, markdown rendering, and progress indicators
• Thinking indicator while the LLM processes
• Permission prompts for tool confirmations

Slash Commands

Inside interactive mode, use / commands to control the session:

Command	Action
/help	Show available commands
/model	Switch models mid-session
/memory	View or add persistent memory
/clear	Clear conversation context
/cost	Show token usage and cost
/exit	End the session

Piped Input

CLIver reads from stdin when data is piped:

cat error.log | cliver "What's causing these errors?"
git diff | cliver "Write a commit message for these changes"

Model Selection

# Use a specific model
cliver -m deepseek-reasoner "Explain this algorithm"

# Set default model
cliver model default minimax-2.7