# TASK: Research OpenClaw Architecture & AI Agent Communities **Priority:** LOW **Agent:** engineer **Filed:** 2026-02-16 **Filed by:** Doug (via Claude Code session) --- ## Goal Research OpenClaw (openclaw.ai) — an open-source personal AI agent by Peter Steinberger. Doug is interested in how it works, what patterns we can steal, and whether it's worth integrating with AgentPi later. Also research the emerging AI agent community — chat rooms, forums, Discord servers where people are building personal AI agents. ## Research Questions ### OpenClaw 1. What is the architecture? How does it run? (Local? Cloud? Hybrid?) 2. How does it communicate? (WhatsApp, Telegram, Signal, SMS?) 3. What's the cron/heartbeat pattern? How does it decide when to do work? 4. What AI model does it use? (Claude API? OpenAI? Both?) 5. What are the key features that make it popular? 6. Peter Steinberger was hired by OpenAI on Feb 15, 2026 — what does that mean for the project? 7. How does it compare to our AgentPi architecture? 8. What ideas can we steal for AgentPi? 9. Is there a risk to using Claude API with OpenClaw? (Anthropic ToS concerns?) 10. GitHub repo: look at the code structure, key files, how jobs/tasks work ### AI Agent Communities 1. What Discord servers, forums, or chat rooms exist for personal AI agents? 2. Are there subreddits or Hacker News threads about this space? 3. What other open-source personal AI agents exist besides OpenClaw? 4. What patterns are emerging in this space? ## Output Write a comprehensive research report to: `/data/Sandbox/AgentPi/research/openclaw_report.md` Structure: 1. Executive Summary (what Doug needs to know in 2 minutes) 2. OpenClaw Architecture Deep Dive 3. Comparison to AgentPi 4. Ideas to Steal 5. AI Agent Community Landscape 6. Recommendations ## DO NOT - Do NOT install or run OpenClaw — research only - Do NOT sign up for any services - Do NOT make changes to AgentPi code in this task