CLI Reference

rho-agent main

The primary command. Starts an interactive REPL or runs a one-shot task.

rho-agent main [PROMPT] [OPTIONS]

Without a prompt argument, the agent starts in interactive mode. With a prompt, it runs the task and exits.

Flags

Flag	Description
--profile <name\|path>	Permission profile: readonly, developer, eval, or path to a YAML file
--backend <name>	Execution backend: local (default) or daytona
--upload <local:remote>	Upload files to sandbox (repeatable, format: ./local:/remote; Daytona only)
--working-dir <path>	Set the agent's working directory
--system-prompt / -s <file.md>	Markdown system prompt file with YAML frontmatter
--var key=value	Set a template variable (repeatable)
--vars-file <file.yaml>	Load template variables from a YAML file
--prompt / -p <text>	Prompt text for one-shot mode
--model <name>	Model to use (default: gpt-5-mini or OPENAI_MODEL)
--base-url <url>	API endpoint override
--config / -c <path>	Agent config YAML file
--output <path>	Write the agent's final response to a file
--auto-approve / -y	Skip tool approval prompts
--resume <id> / -r <id>	Resume a saved conversation (-r latest for most recent)
--list / -l	List saved conversations
--reasoning-effort <level>	Reasoning effort for o1/o3 models

Examples

# Interactive session with developer profile
rho-agent main --profile developer --working-dir ~/proj/myapp

# One-shot task
rho-agent main "find all TODO comments in the codebase"

# Prompt template with variables
rho-agent main --system-prompt examples/job-failure.md \
  --var cluster=prod \
  --var log_path=/mnt/logs/123

# Resume the most recent conversation
rho-agent main -r latest

# Write output to a file
rho-agent main "summarize this project" --output summary.md

rho-agent monitor

Interactive command center for managing running agents.

rho-agent monitor <dir>

Takes a positional argument pointing to a session directory (e.g. ~/.config/rho-agent/sessions).

Monitor commands

Once inside the monitor, these commands are available:

Command	Description
ps	List sessions in the directory
watch <prefix>	Live-follow a session's trace output
cancel <prefix\|all>	Cancel a running session
pause <prefix\|all>	Pause a running session
resume <prefix\|all>	Resume a paused session
directive <prefix> <text>	Inject a directive into a running agent
help	Show all monitor commands
quit	Exit the monitor

rho-agent ps

List sessions in a directory.

rho-agent ps <dir>

Takes a positional argument pointing to a session directory (e.g. ~/.config/rho-agent/sessions).

rho-agent cancel

Cancel running agent sessions.

rho-agent cancel [PREFIX] --dir <dir> [--all]

Flag	Description
--dir <dir>	Session directory (required)
--all	Cancel all running agents

Pass a session ID prefix to cancel a specific agent, or use --all to cancel every running session.

rho-agent export

Export a session trace as ATIF JSON (Agent Trajectory Interchange Format).

rho-agent export <session-id> [OPTIONS]

Flag	Description
--dir <dir>	Sessions directory (default: ~/.config/rho-agent/sessions)
--output / -o <path>	Output file path (default: stdout)

Examples

# Print ATIF JSON to stdout
rho-agent export abc123 --dir ~/.config/rho-agent/sessions

# Write to a file
rho-agent export abc123 -d ~/.config/rho-agent/sessions -o trajectory.json

# Pipe through jq for inspection
rho-agent export abc123 -d ~/.config/rho-agent/sessions | python -m json.tool