Discord Community Manager Agent

🤖 An AI-powered community manager for Discord servers, Telegram groups, and Slack workspaces — answers FAQs, welcomes new members, and keeps conversations on track.

What It Does

This agent acts as a helpful, always-on community manager in your group chats. It monitors conversations, answers frequently asked questions, welcomes new members, and provides helpful resources — all while respecting conversation flow and only jumping in when mentioned or when it can genuinely help.

Perfect for:

• Discord servers with active Q&A channels
• Telegram groups that need FAQ automation
• Slack workspaces with community channels
• Open source projects with support channels
• NFT/crypto communities with repetitive questions

Key Features

✅ Smart FAQ Answering — Pre-loaded knowledge base for common questions
✅ New Member Welcoming — Friendly onboarding messages (when appropriate)
✅ Mention-Only Mode — Only responds when @mentioned in groups (respects conversation flow)
✅ Customizable Knowledge Base — Edit MEMORY.md to teach it about your community
✅ Multi-Platform — Works with Discord, Telegram, Slack, Signal
✅ Lightweight & Fast — Uses Claude Haiku for instant responses
✅ Read-Only in Groups — Safe for public channels, no memory writes

Quick Start

# Install from Hatchery
hatchery run @openclaw/discord-community-manager my-community-bot

# Configure your messaging platform
cd ~/.hatchery/instances/my-community-bot
cp workspace/.env.example .env
nano .env  # Add your TELEGRAM_BOT_TOKEN or DISCORD_BOT_TOKEN

# Customize the knowledge base
nano workspace/MEMORY.md  # Add your FAQs, resources, community info

# Start the agent
hatchery start my-community-bot

Configuration

Required Environment Variables

Choose ONE messaging platform:

# For Telegram
TELEGRAM_BOT_TOKEN=your_bot_token_here
# Get from @BotFather: https://t.me/botfather

# For Discord
DISCORD_BOT_TOKEN=your_bot_token_here
# Get from: https://discord.com/developers/applications

# For Slack
SLACK_BOT_TOKEN=xoxb-your-token-here
SLACK_APP_TOKEN=xapp-your-token-here
# Get from: https://api.slack.com/apps

Customizing Your Knowledge Base

Edit workspace/MEMORY.md to teach the agent about your community:

## Common Questions

### "How do I get started?"
[Your onboarding process here]

### "Where can I find the documentation?"
Docs: https://your-docs-url.com

### "What are the server rules?"
1. Be respectful
2. No spam
3. Stay on topic

The agent will use this knowledge base to answer questions accurately.

Usage Examples

Basic Q&A

User in Discord:

@CommunityBot where can I find the docs?

Agent:

📚 You can find our documentation at https://docs.yourproject.com — the Getting Started guide is a great place to begin!

New Member Welcome

When a new user joins (if your platform supports join events):

Agent:

👋 Welcome to the server, @newuser! Feel free to introduce yourself in #introductions. If you have questions, just @mention me or check out our docs at https://docs.yourproject.com

Handling Unknown Questions

User:

@CommunityBot what's the roadmap for Q3?

Agent:

I don't have specific information about the Q3 roadmap in my knowledge base. @mods might be able to help with that question!

How It Works

Architecture

1. Group Session Only — Agent operates in group session mode (public, read-only)
2. Mention Detection — Only responds when @mentioned (configurable)
3. Knowledge Lookup — Searches MEMORY.md for relevant answers
4. Context-Aware — Reads recent messages to understand context
5. Safe & Sandboxed — No write access, can't execute commands

Session Security

• No workspace writes — Agent can't modify files in group sessions
• No command execution — Can't run shell commands
• Web fetch only — Can fetch URLs to provide information
• No sensitive data — Doesn't store conversation logs

Model Choice

Uses Claude Haiku for:

• ⚡ Fast responses — Sub-second reply times
• 💰 Low cost — Perfect for high-volume community channels
• ✅ Good enough — FAQ answering doesn't need Opus-level reasoning

Customization

Adjusting Personality

Edit workspace/SOUL.md to change the agent's tone:

## Tone & Voice
Friendly and helpful, like a knowledgeable community member who's been
around for a while. Casual but not overly chatty. Uses emojis sparingly.

Adding More Knowledge

Add sections to workspace/MEMORY.md:

## Project Resources
- GitHub: https://github.com/yourorg/yourproject
- Discord: https://discord.gg/yourserver
- Twitter: @yourproject

## Common Issues
### "I can't install the package"
[Troubleshooting steps]

Changing Response Behavior

Edit workspace/AGENTS.md to modify when/how the agent responds:

## When to Respond
- Always: When @mentioned with a direct question
- Sometimes: When you see a FAQ you can answer (even without @mention)
- Never: In off-topic channels, during heated debates

Troubleshooting

Agent Isn't Responding in Groups

Check bot permissions:

• Discord: Bot needs "Read Messages" and "Send Messages" permissions
• Telegram: Bot must be added to the group and given permission to read all messages
• Slack: App needs channels:history, chat:write, app_mentions:read scopes

Verify mention detection:

• Make sure you're @mentioning the bot correctly
• Check hatchery logs my-community-bot for errors

Agent Answering Too Often / Not Enough

Edit hatchery.yaml:

channels:
  groupPolicy:
    requireMention: true    # Set false to allow responses without @mention
    allowlist: ["*"]        # Or restrict to specific group IDs

Knowledge Base Not Working

• Make sure MEMORY.md is well-structured with clear headings
• Use simple, direct language
• Test by asking the agent directly about a topic you added

Rate Limiting / Slow Responses

• Haiku is fast, but check your API quota at https://console.anthropic.com
• If hitting limits, consider upgrading your Anthropic plan

Advanced: Multi-Language Support

Add translations to MEMORY.md:

## FAQs (English)
### "How do I start?"
...

## FAQs (Español)
### "¿Cómo empiezo?"
...

The agent will detect the user's language and respond accordingly.

Best Practices

1. Start conservative — Use requireMention: true initially, relax later if needed
2. Keep knowledge base current — Update FAQs as your project evolves
3. Monitor for accuracy — Check logs to see if the agent is giving good answers
4. Set expectations — Pin a message telling users the bot can help with FAQs
5. Have human fallback — The bot should know when to defer to human mods

Contributing

Found a bug or have a feature request? Open an issue on the OpenClaw GitHub.

Want to improve the template? Submit a PR!

License

MIT License — free to use and modify.

---

Built with OpenClaw — Self-hosted AI agents that actually work.