A 7-agent AI team built on OpenClaw — specialized agents that collaborate via Discord to handle real-world tasks.
| Agent | Role | Specialty |
|---|---|---|
| 🎯 Leader | Squad Lead | Day-to-day coordination, primary human interface |
| 🔀 Orchestrator | Orchestrator | Complex multi-step tasks, delegation |
| 🛠️ Builder | Builder | Coding, development, technical implementation |
| 📚 Scholar | Scholar | Research, analysis, knowledge synthesis |
| 🛡️ Sentinel | Sentinel | Monitoring, alerts, security |
| 🔧 Fixer | Fixer | Creative problem-solving, hacks, workarounds |
| Operator | High-risk operations, edge cases |
- 📧 Email — Read, search, and send via Gmail and iCloud
- 📞 Phone calls — Outbound calls via ElevenLabs + Twilio
- 📅 Calendar — Check and manage Google Calendar + iCloud
- 💻 Code — Build features, fix bugs, deploy
- 🔍 Research — Web search, summarization, analysis
- 📊 Monitoring — Watch for events, alert on conditions
- 💬 Messaging — Discord, WhatsApp, Telegram coordination
- 🖼️ Image generation — Memes and graphics via Gemini
┌─────────────────────────────────────────────────────────┐
│ Discord Server │
│ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ │
│ │ #general│ │ #builder│ │#agent- │ │ #alerts │ │
│ │ │ │ │ │ comms │ │ │ │
│ └────┬────┘ └────┬────┘ └────┬────┘ └────┬────┘ │
└───────┼──────────┼──────────┼──────────┼───────────────┘
│ │ │ │
▼ ▼ ▼ ▼
┌─────────────────────────────────────────────────────────┐
│ OpenClaw Gateway │
│ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ │
│ │ Leader │ │Orchestrator │ │ Builder │ │ Scholar │ │ Sentinel │ │ Fixer │ │
│ └──────┘ └──────┘ └──────┘ └──────┘ └──────┘ └──────┘ │
└─────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ External Services │
│ Gmail │ iCloud │ Twilio │ ElevenLabs │ GitHub │ etc │
└─────────────────────────────────────────────────────────┘
- Node.js 20+
- OpenClaw installed (
npm install -g openclaw) - Discord bot tokens (one per agent)
- API keys for services you want to use
npm install -g openclaw
openclaw initEach agent needs:
- A Discord bot token
- A SOUL.md defining personality
- Optional: dedicated workspace
See agents/ for example configurations.
Copy the example configs and add your API keys:
cp config/example.yaml config/openclaw.yaml
# Edit with your keysopenclaw gateway start- Create a Discord server
- Create a bot application for each agent at discord.com/developers
- Enable Message Content Intent for each bot
- Invite all bots to your server
- Configure channels in
openclaw.yaml
Each agent has a SOUL.md that defines:
- Name and role
- Personality traits
- Communication style
- Boundaries and capabilities
See agents/ for examples.
| Service | Purpose | Config Location |
|---|---|---|
| Gmail | Email read/send | ~/.config/gogcli/ |
| iCloud | Calendar, email | ~/.config/icloud/config.json |
| ElevenLabs | TTS, phone calls | ~/.config/elevenlabs/config.json |
| Twilio | Phone number | ~/.config/twilio/config.json |
| GitHub | Code, issues | gh auth login |
~/.openclaw/
├── config.yaml # Main OpenClaw config
├── agents/
│ ├── main/ # Leader (primary agent)
│ │ └── agent/
│ │ └── SOUL.md
│ ├── proxy/ # Orchestrator agent
│ ├── tech/ # Builder agent
│ ├── yoda/ # Scholar agent
│ ├── boba/ # Sentinel agent
│ ├── babu/ # Fixer agent
│ └── maul/ # Operator agent
└── workspace/ # Shared workspace
├── AGENTS.md # Agent instructions
├── SOUL.md # Primary personality
├── USER.md # User context
├── TOOLS.md # Local tool notes
├── MEMORY.md # Long-term memory
├── HEARTBEAT.md # Periodic task config
└── memory/ # Daily memory files
Agents can run periodic tasks:
- Heartbeats — Regular check-ins (every 30 min) for batched tasks
- Cron jobs — Scheduled tasks with precise timing
Example heartbeat config (HEARTBEAT.md):
## Email Check
Check Gmail for important unread messages
## Calendar Check
Look for upcoming events in the next 24 hoursAgents communicate via:
- Discord channels — Public coordination in shared channels
- Sessions — Direct agent-to-agent messaging
- Spawning — Creating sub-agents for complex tasks
- 🔒 Never commit API keys or tokens
- 🔒 Use environment variables or secure config files
- 🔒 Review agent permissions carefully
- 🔒 Agents have read access to your data — trust accordingly
MIT — do whatever you want with this.
Built by @jhammant