Hermes Agent isn't just a chatbot — it's a framework for building autonomous AI agents that can plan, execute, and improve over time. In this tutorial, you'll build a practical research agent that searches the web and sends you a daily briefing on Telegram.
What You'll Build
By the end of this guide, you'll have an agent that can:
- Search multiple web pages and summarize findings
- Plan and delegate subtasks automatically
- Deliver results to your Telegram account
- Run on a schedule without manual intervention
Step 1: Initialize Hermes
If you haven't installed Hermes yet, run the one-line installer:
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
Then start the setup wizard:
hermes setup
Choose Full Setup to configure everything including API keys and the Telegram bot. If you have an existing OpenClaw installation, the wizard can migrate your settings automatically.
Verify everything is working:
hermes doctor
Step 2: Configure Your Model
Select your LLM provider. For a research agent, models with strong reasoning and large context windows work best:
hermes model
Recommended options:
- Claude Sonnet via Anthropic or OpenRouter — excellent reasoning and summarization
- GPT-4o via OpenAI or OpenRouter — fast and reliable
- Local models via Ollama — completely private, no API costs
Step 3: Set Up the Telegram Gateway
The gateway lets Hermes reach you on Telegram instead of requiring you to stay in the terminal:
hermes gateway setup
- Search for @BotFather in Telegram and create a new bot
- Copy the bot token
- Use @userinfobot to get your Telegram user ID
- Paste both into the Hermes setup wizard
Test it by sending a message to your bot from Telegram.
Step 4: Add Web Search with FireCrawl
To give your agent research capabilities, enable web search. We'll use FireCrawl as an example:
- Sign up at firecrawl.dev and get an API key
- Configure it in Hermes:
hermes config set FIRECRAWL_API_KEY your_key_here
Now test the research capability. Try asking your agent via Telegram:
"Find the latest AI model releases this week and summarize the top 3 with key capabilities and release dates."
You should see Hermes plan multiple tasks, delegate them, search the web, and return a structured summary.
Step 5: Automate with Cron
Set up a daily briefing that runs automatically:
hermes cron add "0 9 * * *" --task "Search for latest AI news, summarize top 5 stories, and send summary to Telegram"
This schedules your research agent to run every morning at 9 AM.
Advanced: Multiple Agent Profiles
You can run multiple independent Hermes agents on the same machine, each with its own config, memory, and skills:
hermes profile create research --clone
hermes profile create coding --clone
Switch between them:
research chat
coding chat
Each profile gets its own Telegram bot, API keys, and memory store.
Deployment Beyond Your Laptop
For 24/7 operation, deploy Hermes to a VPS or serverless platform:
- Modal — serverless, scales to zero
- Daytona — cloud development environments
- Any $5/month VPS — persistent, always-on
Best practices for production:
- Use the container backend
- Set explicit allowlists for gateway access
- Store secrets securely with proper file permissions
- Run updates regularly with
hermes update - Never run as
root
Summary
You now have a fully autonomous research agent that can:
- Search the web using FireCrawl
- Plan and delegate complex research tasks
- Deliver daily briefings via Telegram
- Improve its skills over time through the learning loop
Experiment with different data sources, scheduling intervals, and delivery channels to build an agent that fits your workflow.