Hermes Agent Hands-on Practice
This chapter ties everything together with one complete example: from install to a Telegram personal assistant that reports proactively.
Goal
Build a Hermes assistant that:
- Chats and runs tasks normally in the terminal
- Connects to Telegram for remote control
- Pushes to-dos and a news summary to Telegram every morning at 8 AM
- Remembers your preferences with basic security enabled
Step 1: Install
Step 2: Configure the Model
Or manually:
Put the key in ~/.hermes/.env:
Step 3: First Conversation & Building Memory
Tell it your preferences so it writes them to memory:
These get distilled into USER.md / MEMORY.md and persist across sessions.
Step 4: Tighten Tools & Security
- Disable non-essential high-risk tools
- Keep command approval on; allowlist only trusted read-only commands
- Consider switching risky actions to the Docker backend
Step 5: Connect Telegram
Create a bot via Telegram's @BotFather, get the token, then:
Put the token in .env:
Start the gateway:
Be sure to enable DM pairing + user allowlist so only your account can command it.
Send the bot a message on Telegram:
Step 6: Set Up Daily Proactive Push (Cron)
Define the scheduled task in natural language:
Hermes creates a Cron task that runs unattended and pushes results through the gateway.
Step 7: Distill into a Skill
Once a multi-step flow runs smoothly, solidify it:
Then you can simply:
Verification Checklist
Going Further
- Add a Cron audit task to monitor your server and alert
- Use sub-agents to research multiple sources in parallel
- Connect a second platform (e.g., Slack) for your team
- Browse agentskills.io to install more community skills
Next Steps
- Use Cases — more reproducible patterns
- Multi-Agent — scale up your workflows
- Resources — official docs and community