One AI bot is useful. But what if you need different bots for different purposes? A personal assistant on Telegram, a customer support bot on your website, and a coding helper on Discord — each with its own model, personality, and skills.
OpenClaw supports running multiple instances, and this guide covers everything you need to set them up — whether you're using OpenClaw Launch or self-hosting.
Why Run Multiple Bots?
A single bot with one model and one set of skills is fine for casual use. But as you rely on AI more, you'll want specialized agents:
- Personal assistant — Telegram, fast model (Claude Sonnet), web search, reminders, personal context
- Customer support — Web chat on your site, cost-efficient model (DeepSeek V3), trained on your product docs
- Code helper — Discord, powerful model (GPT-5.3 or Claude Opus), code analysis skills
- Team bot — Shared workspace bot with company knowledge base and internal tools
- Content writer — Telegram or web, creative model, writing and research skills
Each bot gets its own model, skills, system prompt, and platform connection. They don't share context or interfere with each other.
Option 1: Multiple Instances on OpenClaw Launch
The easiest way to run multiple bots. From your dashboard, deploy as many instances as you need:
- Go to openclawlaunch.com and configure your first bot
- Deploy it — takes about 10 seconds
- Go back and configure your next bot with different settings
- Deploy that one too
- Manage all of them from one dashboard
Each instance is fully isolated — its own container, config, AI credits, and platform connection. You can start, stop, restart, and view logs for each one independently.
Cost: Each instance requires its own subscription (Lite at $3/month or Pro at $20/month). AI credits are per-instance, so your personal bot's usage doesn't eat into your support bot's budget.
Option 2: Self-Hosting Multiple Containers
If you're self-hosting, you'll run multiple Docker containers on the same VPS. Each container needs:
- A unique host port (e.g., 18789, 18790, 18791)
- Its own config directory with a separate
openclaw.json - Its own bot token (each Telegram/Discord bot needs a unique token)
- Sufficient memory — remember, each container needs 2GB RAM + 3GB swap
On a 16GB VPS, you can comfortably run 4-5 instances. On 8GB, stick to 2-3.
The main challenge is managing multiple configs and keeping track of which container is which. Use descriptive container names like openclaw-personal, openclaw-support, and openclaw-code.
Example Setup: 3-Bot Configuration
Here's a practical example of three bots, each serving a different purpose:
Bot 1: Personal Assistant (Telegram)
- Platform: Telegram
- Model: Claude Sonnet 4.6 via OpenRouter
- Skills: Web search, memory, calculator, weather
- Use case: Daily tasks, quick questions, personal reminders
- Personality: Friendly, concise, proactive
Bot 2: Customer Support (Web Chat)
- Platform: Web chat embedded on your site
- Model: DeepSeek V3 (cost-efficient for high-volume support)
- Skills: Trained on product documentation, FAQ lookup
- Use case: Answer customer questions 24/7
- Personality: Professional, helpful, always links to docs
Bot 3: Code Helper (Discord)
- Platform: Discord server
- Model: GPT-5.3 or Claude Opus (strongest reasoning)
- Skills: Code analysis, code generation, documentation lookup
- Use case: Code reviews, debugging help, architecture discussions
- Personality: Technical, detailed, shows code examples
Tips for Managing Multiple Bots
- Use descriptive names. "My Bot" tells you nothing. "Personal-Telegram-Claude" tells you everything.
- Match models to use cases. Don't use Claude Opus for simple FAQ answers — it's overkill and expensive. Use DeepSeek V3 or GPT-4.1 Mini for high-volume, simpler tasks. Save the powerful models for complex reasoning.
- Set different budgets. Your personal assistant might use $5/month in AI credits. Your code helper might use $20. Track usage per bot and adjust models if costs are too high.
- Different platforms for different audiences. Personal bots on Telegram (private, always in your pocket). Team bots on Discord (shared access). Customer bots on web chat (no app install required).
- Monitor each bot separately. On OpenClaw Launch, each instance has its own logs and status. Check in on each one periodically to make sure it's performing well.
How Many Bots Do You Need?
Start with one. Get comfortable with how OpenClaw works, how your chosen model responds, and what skills are useful. Then add more as you identify new use cases.
Most users end up with 2-3 bots: one personal, one for work or a project, and maybe one experimental bot to try new models and skills.
Ready to deploy your first bot — or your fifth? Get started on OpenClaw Launch and have your AI agent running in under a minute.