Your First Telegram AI Assistant With Hermes: From /start to Live

Most "Telegram AI bot" tutorials start the same way: spin up a VPS, install Python, fight with BotFather, paste in an API key, pray the process survives a reboot. If you've been through that loop and want something that just stays on, this guide is for you.

Hermes is AgentRoost's always-on AI assistant framework. It auto-provisions a private Telegram bot for you, remembers what you told it three days ago, and can run scheduled check-ins — all without a single token you have to manage yourself.

---

What Hermes Actually Is

Hermes is a persistent AI assistant runtime. The key word is persistent: unlike a chatbot session that forgets everything the moment you close the tab, Hermes keeps conversation context and task state alive indefinitely on dedicated hardware. You talk to it through a private Telegram bot that AgentRoost provisions automatically when you launch the workspace.

A few things it handles out of the box:

• Cross-day memory — ask it to remember your project goals on Monday; follow up on Friday and it still has context.
• Scheduled tasks — configure a morning briefing, a daily digest of your notes, or a periodic check on something you're tracking.
• Inbox and notification routing — feed it alerts from other tools and let it summarize, filter, or forward only the ones that matter.
• Research helpers — give it a standing topic to watch and have it surface new information on a schedule.

What it is not: a workflow automation canvas (that's n8n). Hermes is for assistant-style interaction and background monitoring, not wiring up dozens of app integrations with a visual editor.

---

Before You Start: What You Don't Need

The normal checklist for a Telegram bot looks like this:

1. Open BotFather, create a bot, copy the token.
2. Provision a server or find a hosting plan.
3. Set up an environment, manage API keys (OpenAI, Anthropic, or whichever model you want).
4. Write or deploy the bot code.
5. Keep the process alive — systemd, Docker, supervisor, etc.

With Hermes on AgentRoost, that list collapses to nothing. The bot is provisioned automatically. The LLM credits are already included in your subscription — there is no OPENAI_API_KEY to paste anywhere. The process runs continuously on dedicated hardware. You don't install anything.

---

Launch Your Hermes Instance: Step-by-Step

1. Sign Up and Open the Dashboard

Go to agentroost.app and sign up with email/password, Google, Microsoft, or Discord. Plans start at $19.99/mo; billing begins when you launch your first workspace.

2. Create a New Workspace

From your dashboard, click New Workspace. When the framework picker appears, select Hermes.

Give your instance a name — something descriptive works well here (daily-assistant, research-bot, morning-briefing). The name becomes part of how the workspace is identified in your AgentRoost manager.

3. Connect Telegram in One Click

Once the workspace is provisioning (usually under a minute), you'll see a Connect Telegram button in the AgentRoost manager panel. Click it. This opens the AgentRoost manager bot — send it /start and it will hand you a link to your newly provisioned private bot.

Open that link, send /start to your bot, and you're live. No BotFather session, no token, no webhook URL to configure.

4. Talk to It

Send a message. The AI responds using whatever model is active on your plan — 350+ LLM models are available across tiers, and you can switch models anytime from the workspace settings. Everything goes through included credits; you're not burning a personal API key.

First message tip: Give it a standing context block. Something like: "I'm a freelance developer working on three client projects. My main goals this month are X, Y, Z. Always remember this." Hermes will carry that context forward.

5. Set Up a Scheduled Task (Optional but Powerful)

Inside the workspace settings you can configure scheduled prompts — essentially a cron-style trigger that runs a prompt on a schedule and messages you the result.

Example scheduled tasks that work well:

• Daily standup summary — every morning at 8 AM, Hermes lists what you said you'd do yesterday and asks for an update.
• Research digest — every Friday afternoon, summarize the notes you dropped in during the week.
• Reminder pings — a weekly prompt to review your open loops and surface anything overdue.

The scheduling UI in the manager is a simple form: pick the time, write the prompt, save. No cron syntax, no YAML.

---

What Runs Where

It's worth being concrete about where things live:

Component	Where it runs
Hermes runtime	Dedicated container on AgentRoost hardware
Your Telegram bot	Auto-provisioned, private to your workspace
Conversation memory	Persisted in your workspace volume (your data)
LLM inference	Routed through AgentRoost's included credit pool
Scheduled tasks	Managed by the Hermes scheduler, not your device

Because the runtime is containerized and always on, your assistant responds even when your laptop is closed. Scheduled tasks fire on time even when you haven't opened Telegram that day.

---

Realistic Use Cases

Personal research assistant Tell Hermes the topics you're tracking. Paste in links, article snippets, or notes throughout the week. On Friday, ask it to synthesize what you've accumulated. It remembers everything you dropped in.

Monitoring and alerting companion Use a scheduled task to check a status page or API endpoint and message you only if something is off. Not as powerful as a full n8n workflow for complex alerting, but fast to set up for simple monitoring.

Inbox triage helper Forward key notifications into the Telegram chat, add context, and let Hermes help you decide what needs action versus what can wait.

Team briefing for solo operators Start each day with a /briefing command that Hermes responds to with a structured overview of your current priorities and open tasks (based on what you've told it over the past few days).

---

How This Compares to DIY

If you've self-hosted a Telegram bot before, you know the hidden costs:

• VPS time — even a small instance runs ~$6/month, and that's before you set anything up.
• Model API costs — GPT-4o at any real usage volume adds up fast, and you're managing that separately.
• Uptime babysitting — processes crash, Docker containers need restarting, SSH sessions are involved.
• BotFather friction — every time you rebuild or migrate, you're back in BotFather generating new tokens.

AgentRoost Hermes starts at $19.99/month, all-in. That's the server and the LLM credits and the bot provisioning and zero DevOps overhead bundled together. Plus and Pro tiers add more compute and more included credits as your usage grows.

There's a 14-day money-back guarantee and you can cancel anytime — no annual commitment.

---

Get Started on AgentRoost

If Hermes sounds like the right fit, the fastest path is:

1. See what's included in each plan — the Hermes framework is available on all paid tiers.
2. Read the Hermes overview if you want more detail before committing.
3. Or just launch a workspace and pick Hermes — you'll be chatting with your bot in under two minutes.

The AI credits are already there. The bot is already provisioned. All that's left is the /start.