Command reference

A practical cheat-sheet of the commands I actually use. It is not exhaustive, the authoritative lists are the official CLI Commands and Slash Commands references.

There are two surfaces: terminal commands (hermes ..., run from your shell) and slash commands (/..., typed inside a session).

Terminal commands

Starting and resuming

hermes                       # start the interactive CLI
hermes chat -q "..."         # one-shot, non-interactive
hermes -z "..."              # purest one-shot: final answer only, nothing else on stdout
hermes --tui                 # launch the modern terminal UI
hermes -c                    # resume the most recent session
hermes -c "my project"       # resume by title
hermes -r <id-or-title>      # resume a specific session

Useful hermes chat flags: -m <model>, --provider <name>, -t web,terminal,skills (toolsets), -s <skill> (preload skill), --checkpoints, --yolo, --worktree, --safe-mode (disable all customizations to isolate a problem).

Setup and models

hermes setup                 # configuration wizard
hermes model                 # add providers, run OAuth, set API keys, pick a model
hermes fallback              # manage fallback providers
hermes tools                 # configure enabled tools per platform
hermes auth                  # manage credentials (add/list/remove/status)

hermes model vs /model

hermes model (in your shell) is the full provider setup wizard, it can add providers and run OAuth. /model (inside a session) only switches between providers you have already configured.

Health and maintenance

hermes doctor                # diagnose config, dependencies, service health
hermes update                # pull latest + reinstall (see Updating page)
hermes version               # show version
hermes config check          # list missing config options
hermes config edit           # open config.yaml in your editor
hermes backup                # zip ~/.hermes for backup
hermes import <zip>          # restore a backup
hermes logs                  # view/tail agent + gateway logs
hermes prompt-size           # byte breakdown of the system prompt (runs offline)

State management

hermes sessions list         # browse past sessions
hermes sessions export f.jsonl
hermes profile create coder  # new isolated profile
hermes profile list
hermes curator status        # skill maintenance
hermes checkpoints           # inspect the /rollback shadow store
hermes pairing approve telegram ABC12DEF

Gateway and automation

hermes gateway               # run the messaging gateway
hermes gateway setup         # interactive gateway wizard
hermes gateway install       # install as a systemd/launchd service
hermes gateway status
hermes cron                  # inspect/tick the scheduler
hermes send <platform> "..." # one-shot message to a platform (no agent loop)

Slash commands (in-session)

Session control

Command	Description
/new [name] (/reset)	Fresh session; optional title
/resume [name], /sessions	Resume a named session / browse sessions
/title <name>	Name the current session
/compress [here N | focus topic]	Summarize history to reclaim context
/usage, /insights	Token/cost usage; 30-day analytics
/status	Model, profile, session info + local recap
/rollback [N]	List/restore filesystem checkpoints
/goal <text>	Set a standing goal (Ralph loop)
/stop	Kill running background processes
/quit (/exit)	Exit the CLI

Configuration & tools

Command	Description
/model [name]	Switch among configured models (--global to persist)
/reasoning [level|show|hide]	Reasoning effort and display
/verbose	Cycle tool-progress: off → new → all → verbose
/voice [on|off|tts|status]	CLI voice mode
/yolo	Toggle approval-bypass for the session
/tools, /toolsets	List/enable/disable tools
/skills, /<skill-name>	Manage skills; load any skill as a command
/learn <source>	Distill a reusable skill from a workflow/URL/notes
/curator, /cron, /plugins	Maintenance, scheduling, plugin status
/reload-mcp, /reload-skills, /reload	Reload MCP / skills / .env without restarting

Messaging-only

/sethome (set delivery channel), /approve / /deny (resolve a dangerous-command prompt), /update, /restart, /commands.

Key environment variables

Set these in ~/.hermes/.env (secrets) rather than config.yaml.

Variable	Purpose
OLLAMA_CONTEXT_LENGTH	Context window for the local Ollama server (set to 65536)
HERMES_HOME	Profile/home directory (set automatically by profiles)
HERMES_YOLO_MODE=1	Bypass approval prompts
HERMES_STREAM_READ_TIMEOUT	Streaming read timeout for slow local endpoints
TELEGRAM_BOT_TOKEN, DISCORD_BOT_TOKEN	Gateway bot tokens
TELEGRAM_ALLOWED_USERS, DISCORD_ALLOWED_USERS	Gateway authorization allowlists
API_SERVER_ENABLED, API_SERVER_KEY	OpenAI-compatible API server
GROQ_API_KEY, VOICE_TOOLS_OPENAI_KEY	Optional cloud STT/TTS (local Whisper needs none)

The complete list is in the official Environment Variables reference.

Official catalogs & deeper references

• Model Catalog
• Tools Reference and Toolsets Reference
• Bundled Skills Catalog (~90 skills) and Optional Skills Catalog
• FAQ & Troubleshooting