The gap between “AI assistant” and “AI that works for you” is whether the agent runs when you are not at your desk. Connecting Hermes Agent to Telegram closes that gap. You send a task from your phone. The agent runs it on the server. You get the result in your Telegram chat — whether you sent it at 2pm or 2am.
This guide covers the complete Telegram integration: creating the bot, configuring Hermes, testing the connection, and using it effectively from mobile.
Quick Answer: To connect Hermes Agent to Telegram, create a bot via @BotFather to get a bot token, get your Telegram user ID from @userinfobot, add both to your Hermes
.envfile, and restart the agent. From this point, you control Hermes entirely through Telegram — send tasks, receive results, check status, and manage schedules from any device.

Why Telegram for AI Agent Control
Telegram is the most common Hermes control interface in the community for a few practical reasons:
- Always available. Telegram runs on every device. No browser tab required.
- Bot API is mature. Telegram’s bot API is well-documented and free at any usage scale.
- Access control built-in. User ID whitelisting means only you (and people you authorize) can send commands.
- File handling. Telegram supports sending and receiving files — useful when Hermes returns research documents, reports, or generated content.
- Notification-first. When Hermes completes a long-running task, it pings you with the result. You are not checking a dashboard; the result comes to you.
Hermes also supports Discord, Slack, WhatsApp, and Signal — but Telegram is the most reliable for solo operators running personal workflows.
Prerequisites
Before starting:
- Hermes Agent installed on a VPS (see the VPS installation guide)
- Telegram account on your phone
- SSH access to your VPS
Step 1: Create a Telegram Bot via BotFather
BotFather is Telegram’s official bot management interface. Open Telegram and search @BotFather.
Send:
/newbot
BotFather will ask two questions:
- Name your bot — this is the display name (e.g., “My Hermes Agent”)
- Choose a username — must end in
bot(e.g.,my_hermes_2026_bot)
After completing both, BotFather sends a message with your bot token — a long string that looks like:
1234567890:AAHdqTcvCH1vGWJxfSeofSs0K6oJXGc28v8
Copy this. You will need it in Step 3.
Step 2: Get Your Telegram User ID
The user ID whitelist ensures only your account can send commands to the agent. Search for @userinfobot on Telegram.
Send:
/start
The bot returns your profile info including your numeric user ID — typically 9–10 digits. Copy this.
If you want multiple people to access the same Hermes instance (a team setup), collect all their user IDs at this step.
Step 3: Configure Hermes
SSH into your VPS:
ssh root@YOUR_VPS_IP
cd hermes-agent
nano .env
Find and update these lines:
# Telegram Bot Token from BotFather
TELEGRAM_BOT_TOKEN=1234567890:AAHdqTcvCH1vGWJxfSeofSs0K6oJXGc28v8
# Your Telegram User ID (comma-separated for multiple users)
TELEGRAM_ALLOWED_USERS=987654321
# Optional: For team access, add comma-separated IDs
# TELEGRAM_ALLOWED_USERS=987654321,123456789,555666777
Save: Ctrl+X, Y, Enter.
Restart Hermes:
docker compose restart
Wait 30 seconds for it to reinitialize.
Step 4: Test the Connection
On Telegram, search for your bot’s username (the one you set with BotFather) and open the chat.
Send:
/start
Hermes should respond within 5–10 seconds with a status message confirming it is running and listing available commands.
If no response:
# Check logs for connection errors
docker compose logs hermes | tail -20
Common issues:
- Token has extra spaces — edit
.envand verify no trailing whitespace after the token - User ID is wrong — re-verify via
@userinfobot, copy the exact number - Container not running —
docker compose psto check status
Step 5: Learn the Core Commands
Once connected, these are the commands you will use daily:
General:
/start — Check agent status
/help — List all available commands
/stats — Usage and cost summary
Task management:
[natural language] — Any message is a task
/tasks list — See active and scheduled tasks
/tasks cancel [id] — Cancel a scheduled task
/tasks status [id] — Check a running task
Memory and skills:
/memory search [query] — Search past task history
/memory save [note] — Store a reference note
/skills list — List installed skill packs
/skills install [name] — Install a skill from Skills Hub
Profile management (multi-client):
/profile list — List available profiles
/profile switch [name] — Switch to a client profile
/profile create [name] — Create a new isolated profile
Step 6: Your First Workflow — Morning Digest
Test Hermes’s scheduling by setting up a daily morning briefing. Send via Telegram:
Every morning at 7am, search for the latest AI news and send me a 5-bullet summary in this chat
Hermes will confirm the scheduled task and run it automatically from the next morning. You will receive the digest in your Telegram chat without any further action.
To verify the cron is set:
/tasks list
You should see the morning digest task with its schedule.
Step 7: Mobile Workflow — Sending Tasks on the Go
The practical advantage of Telegram integration is sending tasks from your phone between meetings, during commutes, or before sleep. Some examples:
Research while you wait:
Research the top 5 competitors for [client name] and summarize their recent marketing moves
Hermes runs the research. You get a briefing doc in your Telegram chat within a few minutes.
Content on demand:
Write 3 LinkedIn post ideas for [topic] in a professional but casual tone. Keep each under 150 words.
Quick data lookups:
What were the top AI model releases in the last 7 days?
File-based tasks: Send a PDF or text document directly in the chat, then type:
Summarize this document and extract the 3 most important action items
Hermes processes the attached file and returns the summary.
Setting Up Converse Mode for Sensitive Tasks
For workflows involving emails, publishing, or any action that touches external systems, enable Converse Mode so Hermes shows you the proposed action before executing.
Send:
/mode converse
In Converse Mode, Hermes will send messages like:
📋 Proposed action: Send cold email to [email protected] with this draft:
[email content]
Reply /approve to send or /cancel to stop.
This keeps you in control while still automating the research and drafting steps.
Switch back to autonomous for trusted, non-sensitive workflows:
/mode autonomous
Managing Multiple Clients from One Telegram Chat
If you manage multiple client profiles, you can control all of them from a single Telegram bot by switching profiles:
/profile switch client-acme
Write 3 LinkedIn posts for this week based on their recent product updates
Then:
/profile switch client-beta
Generate a competitive analysis for their Q3 campaign planning
Each profile maintains its own memory and skill context, so outputs stay scoped to the correct client.
Receiving Large Outputs
For tasks that produce long documents — reports, content batches, research briefs — Hermes sends long text directly or as a file attachment depending on length.
To explicitly request a file output:
Generate the weekly content calendar and send it as a text file
Hermes will send a .txt or .md attachment to your Telegram chat, which you can download directly.
Security Checklist for Telegram-Connected Hermes
Before leaving your setup running:
- ✅
TELEGRAM_ALLOWED_USERScontains only your user ID and trusted collaborators - ✅ VPS firewall is configured (only SSH port open)
- ✅ Bot token is not shared anywhere public (GitHub, config files in repos)
- ✅ Converse Mode is enabled for email/publishing workflows
- ✅ You have tested
/tasks listand confirmed no unexpected scheduled tasks
The primary security risk is someone else accessing your bot — the user ID whitelist prevents this as long as you keep the bot token private.
What to Try This Week
With Hermes running 24/7 and connected to Telegram, three workflows worth setting up immediately:
- Morning news digest — 5-bullet AI news summary at 7am daily
- Weekly competitor monitoring — send changes in competitor activity every Monday
- On-demand content drafts — send a topic from your phone, get a draft in your chat within 2 minutes
If you do not have a VPS yet, Hostinger VPS at $5–$8/month is the most cost-effective starting point for running Hermes 24/7. For the AI model access powering these workflows, OpenRouter is the recommended single-key provider — you get Claude, GPT-4o, Gemini, and 300+ other models with pay-per-use pricing, which is ideal for the variable task mix that a Telegram-controlled Hermes runs.
FAQ
Is my Telegram chat with Hermes private?
Yes. Telegram bot chats are private by default. Only accounts in your TELEGRAM_ALLOWED_USERS whitelist can interact with the bot. Telegram itself uses end-to-end encryption for secret chats, though regular bot API messages are encrypted in transit but pass through Telegram’s servers.
Can I use one Hermes instance with multiple Telegram bots?
Not natively in the base setup — one Hermes instance uses one bot token. For multi-client agency setups where each client should have their own bot, you would need separate Hermes instances or a custom routing layer.
What if I lose internet access — do scheduled tasks still run?
Scheduled tasks run on the VPS server, not on your phone. As long as the VPS is online (which is 99.9%+ uptime on any managed provider), scheduled tasks run regardless of your connectivity.
Can Hermes send me proactive alerts?
Yes. You can configure Hermes to monitor conditions and alert you without you asking. Example: “Notify me via Telegram if any of my monitored competitors publish a new blog post.” Hermes runs this check on a schedule and pings you when the condition is met.
How much does the Telegram integration cost?
Telegram’s Bot API is free. The only costs are your VPS ($5–$10/month) and the AI model API calls you use through providers like OpenRouter.
What if I accidentally delete the Telegram chat?
Your Hermes instance and its memory are stored on the VPS — not in Telegram. Delete the chat, start a new one with the bot, and all your previous data is intact. The Telegram chat is just the control interface; the agent’s state lives on the server.