CLI usage

Interactive terminal interface, slash commands, keybindings, and sessions

Starting Hermes

Run hermes from your terminal to open the interactive TUI:

hermes

If you haven’t run setup yet, you’ll be prompted to configure a provider. You can also run the full setup wizard at any time:

hermes setup

The TUI

Hermes uses prompt_toolkit for a full terminal UI:

• Multiline editing — press Meta+Enter (or Esc then Enter) to insert a newline; press Enter to send
• Slash command autocomplete — type / and press Tab to cycle through completions; subcommands and model names also complete
• Ghost-text suggestions — as you type a slash command, the rest appears in dim text
• Streaming tool output — the ┊ activity feed shows each tool call as it runs
• Interrupt and redirect — press Ctrl+C to cancel the agent mid-run; your next message becomes the new instruction

Slash commands

Type /help at any time to see all available commands. Every command resolves its aliases automatically, so /reset and /new both start a new session.

Session commands

Command	Aliases	Args	Description
/new	/reset	—	Start a fresh session (new session ID + history)
/clear	—	—	Clear the screen and start a new session
/retry	—	—	Resend the last user message to the agent
/undo	—	—	Remove the last user/assistant exchange
/title	—	[name]	Set a title for the current session
/compress	—	—	Manually compress conversation context
/rollback	—	[number]	List or restore filesystem checkpoints
/stop	—	—	Kill all running background processes
/background	/bg	<prompt>	Run a prompt in the background
/resume	—	[name]	Resume a previously-named session
/history	—	—	Show conversation history
/save	—	—	Save the current conversation

Configuration commands

Command	Aliases	Args	Description
/model	—	[provider:model]	Show or change the current model
/provider	—	—	Show available providers and current provider
/personality	—	[name]	Set a predefined personality
/skin	—	[name]	Show or change the display skin/theme
/voice	—	[on|off|tts|status]	Toggle voice mode
/reasoning	—	[level|show|hide]	Manage reasoning effort and display
/prompt	—	[text]	View or set a custom system prompt
/verbose	—	—	Cycle tool progress: off → new → all → verbose
/statusbar	/sb	—	Toggle the context/model status bar
/config	—	—	Show current configuration

Tools & Skills commands

Command	Aliases	Args	Description
/tools	—	[list|disable|enable] [name...]	Manage tools
/toolsets	—	—	List available toolsets
/skills	—	[search|browse|inspect|install]	Search, install, and manage skills
/cron	—	[subcommand]	Manage scheduled tasks
/browser	—	[connect|disconnect|status]	Connect browser tools via CDP
/reload-mcp	/reload_mcp	—	Reload MCP servers from config
/plugins	—	—	List installed plugins and their status

Info commands

Command	Aliases	Args	Description
/help	—	—	Show available commands
/usage	—	—	Show token usage for the current session
/insights	—	[days]	Show usage insights and analytics
/platforms	/gateway	—	Show gateway/messaging platform status
/paste	—	—	Check clipboard for an image and attach it

Exit commands

Command	Aliases	Description
/quit	/exit, /q	Exit the CLI

Keyboard shortcuts

Key	Action
Enter	Send message
Meta+Enter / Esc then Enter	Insert a newline (multiline)
Ctrl+C	Interrupt the running agent turn
Tab	Autocomplete slash command or subcommand
↑ / ↓	Navigate input history
Ctrl+L	Clear the screen

Session management

Starting and resetting

Each hermes invocation starts (or resumes) a session. To start completely fresh:

/new

Session IDs are displayed in the welcome banner. You can name a session with /title my-project and resume it later with /resume my-project.

Retrying and undoing

• /retry — resends your last message. Useful when the agent took a wrong turn.
• /undo — removes the last user/assistant exchange from context. The agent won’t see that turn happened.

Context compression

As conversations grow, context usage approaches the model’s limit. Hermes handles this in two ways:

Auto-compression fires automatically when prompt token usage reaches the configured threshold (default: 50% of context length). The middle turns of the conversation are summarized by a fast auxiliary model (default: google/gemini-3-flash-preview), keeping the first three and last four turns intact.

Manual compression via /compress triggers the same process immediately — useful before a complex task to free up space.

Configure compression in ~/.hermes/config.yaml:

compression:
  enabled: true
  threshold: 0.50       # Compress at 50% of context limit
  summary_model: "google/gemini-3-flash-preview"

Filesystem rollback

Hermes takes a filesystem snapshot before destructive file operations (write, patch) once per conversation turn. To list checkpoints:

/rollback

To restore a specific checkpoint:

/rollback 3

Background tasks

Run any prompt in the background without blocking your conversation:

/background run the test suite and summarize failures
/bg deploy the staging branch

The agent spawns the task and returns immediately. Progress notifications appear in your chat as the background process runs. Control verbosity with /verbose or the display.background_process_notifications config key.

Personalities

Personalities apply a predefined system prompt addendum to shape the agent’s tone and style. Switch at any time:

/personality kawaii
/personality concise
/personality technical

Built-in personalities include kawaii, concise, technical, creative, teacher, catgirl, pirate, shakespeare, surfer, noir, uwu, philosopher, and hype. You can define custom personalities in config.yaml under the personalities key.

Skins and themes

The skin system customizes banner colors, spinner faces, tool prefix, response box label, and branding text. Change at runtime:

/skin
/skin ares
/skin mono

Built-in skins

Skin	Description
default	Classic Hermes — gold and kawaii
ares	War-god theme — crimson and bronze, with spinner wings
mono	Monochrome — clean grayscale
slate	Cool blue — developer-focused
poseidon	Ocean-god theme — deep blue and seafoam
sisyphus	Austere grayscale with persistence theming
charizard	Volcanic theme — burnt orange and ember

To set a skin permanently, add it to ~/.hermes/config.yaml:

display:
  skin: slate

Custom skins

Drop a YAML file in ~/.hermes/skins/<name>.yaml. All fields are optional — missing values inherit from the default skin:

name: cyberpunk
description: Neon-soaked terminal theme

colors:
  banner_border: "#FF00FF"
  banner_title: "#00FFFF"
  banner_accent: "#FF1493"

spinner:
  thinking_verbs: ["jacking in", "decrypting", "uploading"]
  wings:
    - ["⟨⚡", "⚡⟩"]

branding:
  agent_name: "Cyber Agent"
  response_label: " ⚡ Cyber "

tool_prefix: "▏"

Activate with /skin cyberpunk.

Voice mode

Hermes supports speech-to-text input and text-to-speech output.

/voice on        # Enable voice input (press Ctrl+B to record)
/voice tts       # Enable TTS for agent responses
/voice status    # Show current voice settings
/voice off       # Disable voice

Voice input uses faster-whisper locally by default (no API key needed). TTS defaults to Microsoft Edge TTS (free). See the Configuration page for provider options.