How to Set Up Multiple OpenClaw Agents on Telegram (Step-by-Step)

What You'll Build

By the end of this tutorial, you'll have:

✅ 3 Telegram bots (Marketing, DevOps, Support)
✅ Each bot connected to a distinct OpenClaw agent
✅ Cross-agent communication (agents can invoke each other)
✅ Shared memory across all agents
✅ Production-ready setup with logging and monitoring

Time required: 30-45 minutes

Prerequisites

Before starting, ensure you have:

1. OpenClaw installed (Installation guide)
2. A Telegram account
3. Basic terminal/command-line knowledge
4. Node.js 18+ (for OpenClaw)
5. A server or local machine where OpenClaw can run 24/7

Step 1: Create Telegram Bots via BotFather

Telegram uses BotFather to manage bots. You'll create 3 bots, one for each agent.

1.1 Open BotFather

1. Open Telegram
2. Search for @BotFather (official Telegram bot)
3. Start a chat: /start

1.2 Create Your First Bot (Marketing Agent)

Send: /newbot

BotFather will ask:

Alright, a new bot. How are we going to call it? 
Please choose a name for your bot.

You reply: Marketing Agent

BotFather:

Good. Now let's choose a username for your bot. 
It must end in `bot`. Like this, for example: TetrisBot or tetris_bot.

You reply: marketing_agent_bot (must be unique across all Telegram)

BotFather responds:

Done! Congratulations on your new bot. You will find it at 
t.me/marketing_agent_bot. You can now add a description...

Use this token to access the HTTP API:
1234567890:ABCdefGHIjklMNOpqrsTUVwxyz1234567890

For a description of the Bot API, see this page: 
https://core.telegram.org/bots/api

Save this token! You'll need it for OpenClaw configuration.

1.3 Repeat for Other Agents

Create 2 more bots:

Bot 2: DevOps Agent

• Name: DevOps Agent
• Username: devops_agent_bot
• Token: (save this)

Bot 3: Support Agent

• Name: Support Agent
• Username: support_agent_bot
• Token: (save this)

Step 2: Set Up Bot Profiles (Optional but Recommended)

Make each bot recognizable with a description and profile picture.

2.1 Set Bot Description

Send to BotFather:

/setdescription

Choose bot: @marketing_agent_bot

Enter description:

I handle marketing strategy, content creation, and SEO for the team. Ask me about blog posts, social media, or campaigns.

Repeat for other bots with their respective descriptions.

2.2 Set Bot Profile Picture

Send to BotFather:

/setuserpic

Choose bot, then upload an image (PNG/JPG, square, 512x512 recommended).

2.3 Set Bot Commands (Auto-Complete Menu)

Send to BotFather:

/setcommands

Choose bot: @marketing_agent_bot

Enter commands:

status - Check agent status
help - Get help with marketing tasks
publish - Publish content
analytics - View marketing metrics

Users will see these commands when they type / in the bot chat.

Step 3: Configure OpenClaw Gateway

OpenClaw uses a central gateway to route messages between channels (Telegram, Discord, etc.) and agents.

3.1 Locate Your OpenClaw Config

OpenClaw config is typically at:

~/.openclaw/config.json

Or in your project directory if you're using a workspace-specific setup.

3.2 Add Telegram Channel Configurations

Edit config.json:

{
  "channels": {
    "telegram": {
      "enabled": true,
      "bots": [
        {
          "name": "marketing",
          "token": "1234567890:ABCdefGHIjklMNOpqrsTUVwxyz1234567890",
          "agent": "marketing-agent",
          "description": "Marketing and content strategy"
        },
        {
          "name": "devops",
          "token": "0987654321:ZYXwvuTSRqponMLKjihGFEdcba0987654321",
          "agent": "devops-agent",
          "description": "Infrastructure and deployments"
        },
        {
          "name": "support",
          "token": "1122334455:AaBbCcDdEeFfGgHhIiJjKkLlMmNnOoPpQqRr",
          "agent": "support-agent",
          "description": "Customer support and issue triage"
        }
      ]
    }
  }
}

Key fields:

• name — Internal identifier for this bot (lowercase, no spaces)
• token — Bot token from BotFather
• agent — Name of the OpenClaw agent this bot routes to
• description — Human-readable description (for logs/UI)

Step 4: Create OpenClaw Agent Configurations

Each Telegram bot needs a corresponding agent in OpenClaw.

4.1 Create Agent Directory Structure

mkdir -p ~/.openclaw/agents/marketing-agent
mkdir -p ~/.openclaw/agents/devops-agent
mkdir -p ~/.openclaw/agents/support-agent

4.2 Define Agent Identity Files

Create ~/.openclaw/agents/marketing-agent/IDENTITY.md:

# Marketing Agent
 
- **Name:** Nadia
- **Role:** Marketing Strategist
- **Vibe:** Creative, data-driven, strategic
- **Emoji:** 📊
 
I handle content strategy, SEO, social media, and campaign planning.

Create ~/.openclaw/agents/marketing-agent/SOUL.md:

# SOUL.md — Marketing Agent
 
You are Nadia, the marketing lead.
 
Your role:
- Content strategy and creation
- SEO optimization
- Social media management
- Campaign planning and execution
 
When users ask marketing questions, provide actionable insights.
When you need DevOps help (deployments, infrastructure), spawn a devops-agent session.

Repeat this structure for devops-agent and support-agent with their own identities.

Step 5: Start the OpenClaw Gateway

5.1 Start the Gateway

openclaw gateway start

Expected output:

✓ Gateway started on port 18799
✓ Telegram: Connected 3 bots
  - marketing-agent → @marketing_agent_bot
  - devops-agent → @devops_agent_bot
  - support-agent → @support_agent_bot
✓ Ready to receive messages

5.2 Verify Bot Status

Open Telegram and message any of your bots:

/status

Expected response:

✅ Agent online
Name: Marketing Agent (Nadia)
Status: Active
Uptime: 2 minutes

Step 6: Test Cross-Agent Communication

Now let's test if agents can talk to each other.

6.1 Message the Marketing Bot

Send to @marketing_agent_bot:

Can you deploy the new blog post?

6.2 What Should Happen

1. Marketing Agent receives your message
2. Recognizes "deploy" = DevOps task
3. Spawns a sub-agent session with DevOps Agent
4. DevOps Agent handles deployment
5. Marketing Agent replies:

✅ Blog post deployed to production.
Live at: https://example.com/blog/new-post

Deployment handled by DevOps Agent.

6.3 Check Gateway Logs

openclaw gateway logs

You should see:

[INFO] Telegram/marketing: User 123456789 → "Can you deploy the new blog post?"
[INFO] Agent marketing-agent: Spawning sub-agent devops-agent
[INFO] Agent devops-agent: Executing deployment task
[INFO] Agent devops-agent: Deployment complete
[INFO] Agent marketing-agent: Reply sent to user

Step 7: Configure Shared Memory

Agents should share context (conversation history, project state).

7.1 Enable Memory Sharing

Edit ~/.openclaw/config.json:

{
  "memory": {
    "enabled": true,
    "backend": "file",
    "path": "~/.openclaw/memory",
    "shared": true
  }
}

7.2 Create Memory Files

mkdir -p ~/.openclaw/memory
touch ~/.openclaw/memory/MEMORY.md

Add initial content to MEMORY.md:

# Shared Memory
 
## Team Context
- Organization: Noqta
- Mission: AI-powered development services
- Current project: Multi-agent Telegram setup
 
## Important Notes
- Marketing Agent handles content
- DevOps Agent handles infrastructure
- Support Agent handles client queries

Now all agents can read/write to this shared memory.

Step 8: Production Deployment

8.1 Run Gateway as a Service

Use PM2 (Node.js process manager):

npm install -g pm2
pm2 start openclaw --name "openclaw-gateway" -- gateway start
pm2 save
pm2 startup

This ensures the gateway restarts automatically on server reboot.

8.2 Set Up Monitoring

pm2 monit

Shows real-time CPU/memory usage, logs, and restarts.

8.3 Log Rotation

pm2 install pm2-logrotate
pm2 set pm2-logrotate:max_size 10M
pm2 set pm2-logrotate:retain 7

Prevents logs from consuming disk space.

Step 9: Security Best Practices

9.1 Environment Variables for Tokens

Don't hardcode bot tokens in config.json.

Use environment variables:

{
  "channels": {
    "telegram": {
      "bots": [
        {
          "name": "marketing",
          "token": "${TELEGRAM_MARKETING_TOKEN}",
          "agent": "marketing-agent"
        }
      ]
    }
  }
}

Set in your shell:

export TELEGRAM_MARKETING_TOKEN="1234567890:ABCdef..."
export TELEGRAM_DEVOPS_TOKEN="0987654321:ZYXwvu..."
export TELEGRAM_SUPPORT_TOKEN="1122334455:AaBbCc..."

9.2 Restrict Bot Access

Use Telegram's allowed users feature:

Edit config.json:

{
  "channels": {
    "telegram": {
      "bots": [
        {
          "name": "marketing",
          "token": "${TELEGRAM_MARKETING_TOKEN}",
          "agent": "marketing-agent",
          "allowedUsers": [123456789, 987654321]
        }
      ]
    }
  }
}

Only users with those Telegram IDs can interact with the bot.

Troubleshooting

Problem: Bot doesn't respond

Check 1: Is the gateway running?

openclaw gateway status

Check 2: Is the bot token correct?

Test manually:

curl https://api.telegram.org/bot<YOUR_TOKEN>/getMe

Should return bot details.

Check 3: Check logs

openclaw gateway logs --tail 50

Problem: Agents can't communicate

Check 1: Are all agents configured?

openclaw agents list

Check 2: Is memory sharing enabled?

Edit config.json:

{
  "memory": {
    "shared": true
  }
}

Check 3: Test sub-agent spawning manually

From within an agent:

sessions_spawn({
  runtime: "subagent",
  agentId: "devops-agent",
  task: "Test task",
  mode: "run"
})

Problem: High memory usage

Solution: Limit session history

Edit agent AGENTS.md:

## Session Management
- Keep last 100 messages only
- Clear old sessions after 24 hours

Next Steps

Now that you have a working multi-agent Telegram setup:

✅ Add more agents (QA, Analytics, etc.)
✅ Customize agent personalities (edit SOUL.md)
✅ Build workflows (marketing → devops → QA pipelines)
✅ Monitor performance (track response times, success rates)

Resources

• OpenClaw Docs: https://docs.openclaw.ai
• Telegram Bot API: https://core.telegram.org/bots/api
• Community Discord: https://discord.com/invite/clawd
• GitHub: https://github.com/openclaw/openclaw

Need help? Book a consultation: /ai-automation