Hermes Agent CLI cheat sheet — commands, flags, and slash shortcuts

Hermes Agent from Nous Research is a model-agnostic, tool-using assistant you run locally or on a VPS.

Hermes does not lock you into one surface. You can use

• the classic hermes / hermes chat CLI,
• the full-screen hermes --tui session,
• a long-running hermes gateway for Telegram, Discord, Slack, and other messaging platforms,
• hermes dashboard for a local browser UI when the web extra is installed.

Those paths share the same config and data under ~/.hermes; this page lists shell commands that matter across those modes.

Below is a dense command reference grouped by task.

Install Hermes Agent and first-run CLI commands

For install and troubleshooting, start with Hermes AI Assistant — Install, Setup, Workflow, and Troubleshooting.

The installer pulls the repo, sets up a Python environment, and wires the hermes executable. After source ~/.bashrc or ~/.zshrc, your default entry point for interactive chat is simply hermes (same family as hermes chat).

Command	Description
`curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh \	bash`
hermes / hermes chat	Start interactive chat after install (default daily entry).
hermes --version / hermes version	Print version information.
hermes completion bash \	zsh \
hermes update [--check] [--backup] [--restart-gateway]	Pull latest code, reinstall deps, optional pre-update home snapshot or gateway restart.
hermes uninstall [--full] [--yes]	Remove Hermes; optional full data deletion.

Native Windows is not supported; use WSL2. Android installs via Termux follow a dedicated path in the upstream docs.

Global flags for every hermes invocation

These flags apply before subcommands and change which profile, which session, or how much personal config loads.

Flag	Description
--profile, -p	Select Hermes profile for this run (overrides sticky default from hermes profile use).
--resume, -r	Resume a session by ID or title.
--continue [name], -c	Continue the latest session, or latest matching a title.
--worktree, -w	Start in an isolated Git worktree for parallel agents.
--yolo	Bypass dangerous-command approval prompts (use with care).
--pass-session-id	Include session ID in the system prompt.
--ignore-user-config	Skip ~/.hermes/config.yaml (defaults only); .env still loads.
--ignore-rules	Skip auto-injection of AGENTS.md, SOUL.md, .cursorrules, memory, preloaded skills.
--tui	Launch the TUI (HERMES_TUI=1 equivalent).
--dev	With --tui, run TS sources via tsx for TUI development.

Isolated automation often pairs hermes chat --ignore-user-config --ignore-rules with hermes -z for reproducible one-shots.

hermes chat, one-shot prompts, and hermes -z

Command / pattern	Description
hermes chat	Interactive or scripted chat; main surface for -q, -m, --provider, toolsets, resume, worktree, checkpoints.
hermes chat -q "..."	One-shot prompt (non-interactive); keeps richer output than -z when tools run.
hermes -z "..."	Scripted one-shot — final answer only on stdout, no banner or session noise. Same agent and tools; best for pipes and scripts.
hermes chat --quiet, -Q	Quieter programmatic mode (banner and tool previews suppressed).
-m / --model, --provider	Per-run model and provider overrides; env HERMES_INFERENCE_MODEL / HERMES_INFERENCE_PROVIDER mirror behavior.
-t / --toolsets	Enable comma-separated toolsets for the run.
-s / --skills	Preload skills (repeat or comma-separated).
--image path	Attach a local image to a single query.
--checkpoints	Enable filesystem checkpoints before destructive edits.
--max-turns N	Cap tool-calling iterations per turn (default from config).
--source	Session source tag (cli vs tool for integrations).

Hermes model outside the session vs /model inside it — Running hermes model from the shell is where you add providers, keys, and OAuth. Slash /model only switches among already configured providers. If you only see OpenRouter in /model, exit the session and complete hermes model.

Model picker, credential pools, and fallback providers

Command	Description
hermes model	Interactive provider and model picker; keys, OAuth, custom endpoints.
hermes auth	Credential pools — add, list, remove, reset for rotation-friendly keys and OAuth.
`hermes fallback [list \	add \
{% raw %}`hermes setup [model \	tts \

Deprecated {% raw %}hermes login / hermes logout — use hermes auth and hermes model instead.

Picking local OpenAI-compatible endpoints versus hosted APIs for hermes model sits on the same trade-offs as general LLM hosting (latency, cost, ops).

Config files and hermes config commands

Configuration resolves as CLI overrides → config.yaml → .env → defaults. API keys belong in .env; structured settings in config.yaml.

Command	Description
hermes config show	Display effective configuration.
hermes config edit	Open config.yaml in $EDITOR.
hermes config set key value	Set values (secrets routed to .env, non-secrets to YAML).
hermes config path / hermes config env-path	Print paths to config and env files.
hermes config check	Detect missing or stale settings.
hermes config migrate	Apply newly introduced options interactively.

Where files live — Everything sits under HERMES_HOME (default ~/.hermes) for config, secrets, memories, skills, sessions, gateway state, and logs.

Session management and hermes profile

Command	Description
hermes sessions list	List recent sessions.
hermes sessions browse	Interactive picker with search and resume.
hermes sessions export	Export sessions (e.g. JSONL).
hermes sessions delete, prune, rename, stats	Delete one session, prune old ones, rename titles, show store stats.
hermes profile list \	use \
hermes profile export / import	Archive or restore a profile tarball.
hermes profile alias	Short wrapper scripts for fast profile switching.

Use hermes -p work chat -q "..." for ad hoc runs without changing the sticky default profile.

Skills hub, toolsets, shell hooks, and plugins

For profile-first configuration and skills tuned to real production workflows by role, see Hermes AI Assistant Skills for Real Production Setups.

Command	Description
hermes tools	Interactive per-platform tool enablement; --summary prints current choices.
hermes skills browse, search, inspect, install, list, check, update, audit, uninstall, publish, snapshot, tap, config	Skills hub workflows including registries and URL installs.
hermes curator status, run, pause, pin, rollback, …	Background skill maintenance and safe rollback.
hermes hooks list, test, revoke, doctor	Declared shell hooks and allowlists in config.
hermes plugins	Composite UI or subcommands to install, enable, disable, remove plugins.

Built-in memory and hermes memory providers

Built-in MEMORY.md / USER.md stay active; external providers add optional recall layers. For how that architecture behaves in practice, read Hermes Agent Memory System — How Persistent AI Memory Actually Works. To compare external backends and activation trade-offs, see Agent Memory Providers Compared — Honcho, Mem0, Hindsight, and Five More.

Command	Description
hermes memory setup	Interactive external memory provider configuration.
hermes memory status	Show active provider settings.
hermes memory off	Disable external provider; built-in files remain.

When a provider is active it may register extra provider-specific top-level subcommands — run hermes --help to see what is wired today.

Messaging gateway, DM pairing, and platforms

Command	Description
hermes gateway setup	Interactive messaging platform setup.
hermes gateway run	Foreground gateway (recommended on WSL, Docker, Termux).
hermes gateway start \	stop \
hermes gateway install \	uninstall
hermes pairing list \	approve \
hermes whatsapp	WhatsApp bridge pairing flow.
hermes slack manifest	Generate Slack app manifest with gateway slash parity.

On WSL, hermes gateway run inside tmux is the resilient pattern when gateway start misbehaves.

Cron scheduler, webhooks, and Kanban

Command	Description
hermes cron …	Create, edit, pause, resume, run, remove scheduled prompts (tick for manual scheduler pass).
hermes webhook subscribe, list, remove, test	Dynamic webhook routes for event-driven runs.
hermes kanban …	Multi-profile task board backed by SQLite; dispatch drives workers.

hermes doctor, logs, backup, and usage insights

Command	Description
hermes doctor [--fix]	Interactive diagnostics and optional auto-repair.
hermes status [--all] [--deep]	Concise status; deeper checks when needed.
hermes dump [--show-keys]	Paste-friendly setup summary for Discord or GitHub issues.
hermes debug share	Upload redacted debug bundle to a paste service (or --local).
`hermes logs [agent \	errors \
{% raw %}hermes backup, hermes import	Zip snapshots of home data and restore paths.
hermes insights [--days N] [--source …]	Token, cost, and activity analytics.

When something breaks after an upgrade, hermes doctor, hermes status, and hermes logs errors -f form the fastest triage loop.

MCP, ACP, web dashboard, and OpenClaw migration

Command	Description
hermes mcp serve	Run Hermes as an MCP server.
hermes mcp add, remove, list, test, configure	Manage MCP client connections from Hermes.
hermes acp	Agent Client Protocol stdio server for editors (extra install may apply).
hermes dashboard [--port …] [--host …]	Local web dashboard (pip install hermes-agent[web]).
hermes claw migrate …	Migrate OpenClaw-style configs into Hermes (--dry-run, presets, optional secrets).

OpenClaw migration — hermes claw migrate reads legacy OpenClaw home directories; for what that stack looked like before moving, see the OpenClaw case study.

Slash commands in the Hermes CLI session

Type / for autocomplete. Commands are case-insensitive; skills register extra /skill-name routes. The tables below are a curated subset; for the full registry see Official Hermes Agent documentation at the end of this article.

Session flow, background tasks, and goals

Command	Description
/new, /reset	New session ID and history.
/resume [name]	Resume a named session.
/compress [focus]	Manual context compression with optional focus topic.
/retry, /undo	Retry last turn or drop last exchange.
/title …	Name the session for later /resume.
/background …, /queue …, /steer …	Parallel background run, queued next prompt, mid-loop nudge after next tool.
/goal …	Persistent multi-turn objective with judge loop (status, pause, resume, clear).
/branch, /fork	Branch the conversation for alternate exploration.

Models, tool toggles, skills, and reload

Command	Description
/model … [--global]	Switch models among configured providers; --global persists default.
/tools …, /toolsets	Session tool toggles and toolset listing.
/skills …	Search, install, and manage skills from chat.
/cron …	Scheduled tasks UI from the CLI session.
/reload-mcp	Reload MCP servers from config.
/reload	Reload .env into the running session without restart.

Usage, help, and quitting

Command	Description
/usage, /insights	Token and cost visibility; analytics snapshot.
/help, /quit	Help or exit the CLI.

Messaging apps (Telegram, Discord, Slack, and others) expose an overlapping slash set plus /approve, /restart, /commands, and related gateway-only helpers — platform differences are documented in the slash command reference under Official Hermes Agent documentation below.

More useful reading

Related pages on this site (broader context for Hermes and terminal agents):

• AI Systems — Self-Hosted Assistants, RAG, and Local Infrastructure — cluster overview and how assistants fit the stack
• AI Systems Memory — memory hub and adjacent guides
• AI Developer Tools — terminal and IDE tooling landscape
• OpenCode Quickstart — another terminal-first agent for ergonomic comparison

Official Hermes Agent documentation

Upstream documentation on hermes-agent.nousresearch.com:

• Hermes Agent
• CLI commands reference
• Slash commands reference

---

Tip. Keep hermes dump and hermes doctor --fix in muscle memory — they turn vague "something broke" reports into actionable diffs against a known-good setup.