Commands
Complete reference for AGTerm CLI commands.
Command Syntax
Commands start with / and are processed by the AGTerm agent:
/command [arguments] [--flags]Core Commands
Help & Status
| Command | Description |
|---|---|
/help, /h | Show available commands |
/status, /s | Display current session status (provider, model, memory) |
/ask <question> | Ask Calliope about AGTerm features |
/tour | Interactive walkthrough of AGTerm capabilities |
Exit
| Command | Description |
|---|---|
/exit, /quit, /q | Exit AGTerm session |
Provider & Model Commands
Switching Providers
/provider [name]Without arguments, shows current provider. With a name, switches to that provider:
/provider groq
# Now using: groqModel Selection
/model [name]Show or set the current model:
/model claude-opus-4-20250514
# Model set to: claude-opus-4-20250514List Models
/models [provider]List available models for current or specified provider:
/models groq
# Available models for groq:
# - llama-3.3-70b-versatile
# - llama-3.1-8b-instant
# - mixtral-8x7b-32768Enable/Disable
/enable <provider|model>
/disable <provider|model>Toggle provider or model availability for the session.
Configuration Commands
View Configuration
/configDisplay current configuration values.
Set Configuration
/config set <key> <value>Set a configuration value for the current session:
/config set max_tokens 4096Save Configuration
/config save [global]Persist configuration. Use global to save to user-level config:
/config save # Save to project
/config save global # Save to ~/.agterm/configReset Configuration
/config resetReset to default configuration values.
Output & Style
Fancy Mode
/fancy [on|off]Toggle enhanced output formatting (colors, animations, ASCII art):
/fancy off # Minimal output for scripting
/fancy on # Rich terminal outputPersona Selection
/persona [name]
/tone [name]Switch the agent’s communication style:
| Persona | Description |
|---|---|
calliope | Poetic, creative with technical precision |
professional | Clear, concise, thorough |
minimal | Extremely brief responses |
/persona minimal
# Switching to minimal personaMemory Management
View Memory Stats
/memoryDisplay token usage and memory statistics:
Memory Status:
- Messages: 24
- Tokens: 12,450 / 50,000
- Compression: 0 roundsCompact Memory
/compactCompress conversation history to free up context space. Older messages are summarized.
Add Context
/context <file>Add file contents to the conversation context:
/context src/main.tsView History
/history [n]Show the last n messages (default: 10):
/history 5Forget Messages
/forget [n]Remove the last n messages from memory:
/forget 3Clear History
/clear, /cClear the entire conversation history.
Session Management
Save Session
/save [name]Save current session state (memory, configuration):
/save my-project
# Session saved as: my-projectResume Session
/resume [name]Load a previously saved session:
/resume my-project
# Resumed session: my-project (24 messages)List Sessions
/sessionsShow all saved sessions:
Saved Sessions:
- my-project (2024-01-15, 24 messages)
- bugfix-auth (2024-01-14, 12 messages)Continue Recent
/continueResume the most recently saved session.
Sub-agent Commands
Direct Sub-agent Calls
/claude <prompt>
/gemini <prompt>
/codex <prompt>Run a prompt with a specific agent:
/claude Explain this authentication flow and suggest improvementsSet Default Sub-agent
/subagent <name>Set the default sub-agent for future calls:
/subagent claudeBackground Jobs
Submit Background Job
/bg <prompt>
/batch <prompt>
/background <prompt>Run a task in the background, allowing you to continue working:
/bg Run all tests and generate a coverage report
# Job submitted: abc123List Jobs
/jobsShow all background jobs and their status:
Background Jobs:
- abc123: running (2m 15s)
- def456: completed (output available)
- ghi789: failed (error logged)Get Job Output
/job <id>Retrieve the output of a completed job:
/job abc123Cancel Job
/cancel <id>Stop a running background job:
/cancel abc123Autonomous Loops
Start Loop
/loop "<prompt>" [--max-iterations N] [--completion-promise "text"]Start an autonomous iteration loop where the agent works toward a goal:
/loop "Fix all failing tests" --max-iterations 20 --completion-promise "All tests passed"Options:
--max-iterations N— Maximum iterations before stopping (default: 50)--completion-promise "text"— Stop when output contains this text
Cancel Loop
/cancel-loopStop an active autonomous loop.
Command Examples
Quick Model Switch for Fast Responses
/provider groq
/model llama-3.3-70b-versatile
What's the time complexity of quicksort?Save Work Before Disconnect
/save afternoon-session
/exitBackground Test Run
/bg npm test -- --coverage
/jobs # Check status
/job abc123 # Get results when doneAutonomous Bug Fixing
/loop "Run tests, fix any failures, repeat until all pass" --max-iterations 30Context-Heavy Analysis
/context src/auth/
/context docs/security.md
Analyze the authentication system for potential vulnerabilities