CLI Reference

hugin create

Interactive wizard for creating new agents.

hugin create

Guides you through:

Naming your agent
Selecting an LLM model
Choosing tools
Creating config, task, and template files

hugin run

Run an agent with a task.

hugin run [options]

Option	Description
`-t, --task`	Task name to run
`-p, --task-path`	Path to agent directory
`-c, --config`	Config name (default: first found)
`--parameters`	JSON string of task parameters
`--max-steps`	Maximum steps (default: 100)
`--storage-path`	Where to save execution data
`--model`	Override LLM model
`--monitor`	Also start the web dashboard
`-i, --interactive`	Run the agent inside the interactive TUI

Examples:

# Interactive mode - prompts for task and directory
hugin run

# Run specific task
hugin run -t my_task -p ./my_agent

# With parameters
hugin run -t analyze -p ./agent --parameters '{"input": "data.csv"}'

# Using local Ollama model
hugin run -t hello -p ./agent --model ollama:llama3.2

# Run with monitoring dashboard
hugin run -t my_task -p ./agent --monitor

# Run inside the interactive TUI
hugin run -t my_task -p ./agent -i

hugin interactive

Open the interactive TUI for browsing and exploring sessions and agents without running anything.

hugin interactive [options]

Option	Description
`-p, --task-path`	Path to agent directory
`-s, --storage-path`	Path to agent storage

Examples:

# Browse all sessions in default storage
hugin interactive

# Browse sessions for a specific agent directory
hugin interactive -p ./my_agent

# Browse a specific storage location
hugin interactive -s ./data/my_agent

hugin monitor

Web dashboard for watching agent execution.

hugin monitor [options]

Option	Description
`-s, --storage-path`	Path to agent storage
`-p, --port`	Server port (default: 8000)
`--no-browser`	Don't auto-open browser

Example:

# Start monitor for specific storage
hugin monitor -s ./data/my_agent

# Custom port
hugin monitor -s ./storage -p 8080

Features:

Real-time interaction flow visualization
Tool call inspection
Agent state browsing
Session/agent selection

hugin rate

Rate an artifact as a human reviewer.

hugin rate [options]

Option	Description
`-s, --storage-path`	Path to agent storage
`--artifact-id`	UUID of the artifact to rate
`--rating`	Rating from 1 (poor) to 5 (excellent)
`--comment`	Optional comment explaining the rating

When run without flags, enters interactive mode: lists available artifacts, prompts for selection, rating, and optional comment. Provide all flags for non-interactive/scripted usage.

Examples:

# Interactive mode - browse and rate artifacts
hugin rate -s ./storage

# Non-interactive mode
hugin rate -s ./storage --artifact-id abc123 --rating 4 --comment "Very useful"

Interactive TUI

Terminal UI for browsing and managing agents. Access it via hugin interactive (browse only) or hugin run -i (run agent inside TUI).

# Browse existing sessions
hugin interactive --storage-path ./storage

# Run an agent inside the TUI
hugin run -t my_task -p ./my_agent -i

Key	Action
`↑/↓` or `j/k`	Navigate
`Enter`	Select / Drill down
`q`	Back / Quit
`r`	Refresh
`N`	New agent (on sessions screen)
`?`	Help

Screens:

Sessions - List all saved sessions
Agents - Agents within a session
Interactions - Agent's execution stack
Detail - Full interaction data

hugin apps

List available example apps.

hugin apps

hugin app

Run a specific app.

hugin app <name> [args]

Example:

hugin app data_analyst
hugin app rap_machine --random-agents

hugin install-models

Install recommended Ollama models.

hugin install-models