Enter passcode to access materials
Contact the instructor for access
OpenClaw Agent Setup & Agentic Systems — Session Materials
Google Drive — Full Session Video
Complete OpenClaw setup walkthrough, start to finish
Open-source agent engine running on your device. Orchestrates tools, memory, and LLM calls.
Telegram · WhatsApp · Slack · Discord
Claude · GPT · Gemini · Local
Calendar, email, browser, APIs…
* relative capability scores, illustrative
Install skills like npm packages. Each module extends your agent's capabilities.
Receives task → selects browser tool → executes → returns result
Click, type, navigate live sites
Sites → structured data
Scalable cloud sessions
Browser agent primitives
| Feature | Emergent (Cloud) | Self-Hosted |
|---|---|---|
| Setup time | Minutes | Hours |
| Data stays local | ||
| Persistent 24/7 | ||
| Local models | ||
| Full control | ||
| Monthly cost | Usage-based | Hardware only |
Never use your primary Gmail on untested agents. Isolated accounts limit blast radius.
Grant only what's needed. Use read-only keys where possible. Revoke after testing.
Every skill runs with your API keys. Treat it like a browser extension — read the code.
Do not port-forward OpenClaw to the internet. Use Telegram as the only interface.
Stanford research showing believable AI agents that plan, reflect, and coordinate in a simulated world. The foundational paper for agent memory and autonomous behavior.
arxiv.org/abs/2304.03442 →How multiple AI agents form networks, share information, and develop emergent behaviors — directly relevant to building agent twins and multi-agent systems.
arxiv.org/abs/2602.19810 →