CLI Essential Commands Every User Should Know

The Hermes CLI at a Glance

The Hermes Agent command-line interface is your primary way to interact with the agent. While you can connect Hermes to Telegram or Discord for mobile access, the CLI offers the most powerful and flexible experience.

After installing Hermes, open your terminal and type:

hermes --help

Hermes Agent v0.8 — Self-improving AI Agent

USAGE:
  hermes [OPTIONS] [MESSAGE]

OPTIONS:
  -m, --model        AI model to use
  -s, --session       Session to use or create
  -c, --continue             Continue last session
  -f, --file          Include file(s) in context
  -p, --pipe                 Read from stdin
  -o, --output        Write response to file
  -v, --verbose              Show detailed logs
  --no-confirm               Skip confirmation prompts
  --json                     Output responses as JSON
  --version                  Show version
  --help                     Show this help

COMMANDS:
  config                     Manage configuration
  session                    Manage sessions
  skill                      Manage skills
  memory                     Manage long-term memory
  cron                       Manage scheduled tasks

Basic Usage Patterns

Interactive Mode (Default)

Just type hermes to start an interactive conversation:

hermes

You'll enter a REPL where you can type messages and receive responses in real-time.

One-Shot Mode

Send a single message and get the response without entering interactive mode:

hermes "What is the capital of France?"

This is perfect for quick questions or scripting. The output goes to stdout, so you can pipe it:

hermes "Generate a random UUID" | pbcopy  # macOS
hermes "Generate a random UUID" | xclip   # Linux

File Context Mode

Include files in your conversation for analysis or modification:

# Include a single file
hermes -f src/app.ts "Add error handling to all routes"

# Include multiple files
hermes -f src/app.ts -f src/routes.ts -f package.json "Review this project"

# Include all files matching a pattern
hermes -f "src/*/.ts" "Find potential memory leaks"

Pipe Mode

Pipe data from other commands directly into Hermes:

# Analyze logs
cat /var/log/nginx/error.log | hermes -p "Summarize the errors from the last hour"

# Explain code
cat src/parser.ts | hermes -p "Explain this code to a junior developer"

# Process data
curl -s https://api.example.com/data | hermes -p "Convert this JSON to a CSV table"

# Git diff review
git diff HEAD~3 | hermes -p "Review these changes and suggest improvements"

Session Management

Sessions are persistent conversation threads. Each session maintains its own history, context, and state.

Essential Session Commands

# Start with a named session
hermes -s my-project

# Continue the last session
hermes -c

# List all sessions
hermes session list

# Show session details
hermes session info my-project

# Delete a session
hermes session delete old-project

# Export session history
hermes session export my-project > session-log.md

In-App Session Commands

While in interactive mode, use slash commands:

> /session new api-redesign
/session switch my-project
/sessions
/session rename api-redesign api-v2
/session delete api-v2

Session Best Practices

One session per project — Keeps context focused

Descriptive names — blog-migration not session-1

Clean up regularly — Delete completed sessions to save disk space

Export important sessions — Save logs before deleting

Configuration

Hermes uses a YAML configuration file at ~/.hermes/config.yaml. You can edit it directly or use the CLI:

# View all configuration
hermes config list

# Get a specific value
hermes config get default_model

# Set a value
hermes config set default_model anthropic/claude-3.5-sonnet

# Reset to defaults
hermes config reset

Key Configuration Options

# ~/.hermes/config.yaml
api_key: sk-or-v1-your-key-here
provider: openrouter
default_model: anthropic/claude-3.5-sonnet
temperature: 0.7
max_tokens: 4096

# Behavior
auto_confirm: false           # Auto-confirm file changes
verbose: false                # Show detailed logs
theme: dark                   # Terminal color theme

# Sessions
session_dir: ~/.hermes/sessions
max_session_history: 100      # Messages to keep in memory

# Memory
memory_enabled: true
memory_dir: ~/.hermes/memory

# Model aliases
model_aliases:
  fast: openai/gpt-4o-mini
  smart: anthropic/claude-3.5-sonnet

For a deep dive, see our Configuration Guide.

Skills Management

Skills extend Hermes' capabilities. Think of them as plugins that teach Hermes new tricks.

# List installed skills
hermes skill list

# Search community skills
hermes skill search "web scraping"

# Install a skill
hermes skill install github-pr-review

# Remove a skill
hermes skill remove github-pr-review

# Update all skills
hermes skill update --all

In-App Skill Commands

> /skills                     # List active skills
/skill info web-search      # Show skill details
/skill toggle web-search    # Enable/disable a skill

Memory Management

Hermes' long-term memory persists across sessions. Use it to store preferences, facts, and project context.

# Save a memory
hermes memory save "I prefer tabs over spaces"

# List all memories
hermes memory list

# Search memories
hermes memory search "deployment"

# Delete a specific memory
hermes memory delete 42

# Clear all memories
hermes memory clear

Power-User Shortcuts

Chaining Commands

Use one-shot mode for automated workflows:

# Generate, then run
hermes "Generate a Python script that checks if port 8080 is open" -o check_port.py && python check_port.py

# Analyze and fix
hermes -f src/broken.ts "Find and fix the bug" --no-confirm

# Batch processing
for file in src/*.ts; do
  hermes -f "$file" "Add JSDoc comments to all functions" --no-confirm
done

Output Formatting

Control output format for scripting:

# JSON output for parsing
hermes --json "List 5 random city names"
# Output: {"cities": ["Tokyo", "Berlin", ...]}

# Write directly to file
hermes -o report.md "Generate a project status report for the web-scraper session"

# Suppress thinking blocks
hermes --quiet "What time is it in Tokyo?"

Environment Variables

Set defaults via environment variables:

export HERMES_MODEL=anthropic/claude-3.5-sonnet
export HERMES_API_KEY=sk-or-v1-your-key
export HERMES_SESSION=default
export HERMES_VERBOSE=true

Add these to your .bashrc or .zshrc for persistence.

Keyboard Shortcuts (Interactive Mode)

| Shortcut | Action |

|----------|--------|

| Ctrl+C | Cancel current response |

| Ctrl+D | Exit Hermes |

| Ctrl+L | Clear screen |

| Up/Down | Navigate message history |

| Tab | Autocomplete commands |

| Ctrl+R | Search message history |

Debugging and Troubleshooting

Verbose Mode

Add -v to see detailed request/response logs:

hermes -v "Hello"

[DEBUG] Connecting to OpenRouter...
[DEBUG] Model: anthropic/claude-3.5-sonnet
[DEBUG] Sending 1 message (245 tokens)
[DEBUG] Response received in 1.2s (312 tokens)
Hello! How can I help you today?

Connection Testing

hermes --test-connection
# ✅ Provider: OpenRouter
# ✅ Model: anthropic/claude-3.5-sonnet
# ✅ API Key: Valid
# ✅ Latency: 142ms

Common Issues

"Model not found" error:

# Check available models
hermes config get provider
hermes --list-models | grep claude

Slow responses:

# Switch to a faster model
hermes config set default_model openai/gpt-4o-mini

Context too long:

# Start a fresh session
hermes session new fresh-start

Next Steps

Now that you're comfortable with the CLI, explore:

• Configuration deep dive — Customize every aspect of Hermes
• Your first conversation — Essential conversation techniques
• Cron scheduling — Automate tasks with scheduled runs

---

Last updated: April 15, 2026 · Hermes Agent v0.8