March 11, 2026 | luma.com/lc69v262 Presented by: Zeugh (@theZeugh)
Your AI agent with real tools
Run Claude/GPT with actual computer access Built-in tools: files, shell, browser, messaging Multi-channel: Telegram, Discord, WhatsApp Self-hosted: You control the data Real automation, not just chat
"Claude Desktop for your server, with superpowers"
ChatGPT/Claude.ai: Can't access YOUR files Can't run YOUR commands Resets every conversation
OpenClaw: Persistent memory Real filesystem access Actual automation
Claude Desktop: Only works on your laptop Can't receive messages when closed Single agent only
OpenClaw: Runs 24/7 on your server Receives Telegram messages Multi-agent coordination
npm install -g openclaw openclaw init openclaw gateway start openclaw webchat
That's it! 5 minutes to a running agent.
You: "Remind me to call mom in 2 hours"
Agent: Creates cron job, sends notification later
You: "Did I go to the gym this week?"
Agent: Checks memory files, accountability report
You: "Build a REST API for a todo app"
Agent: Creates files, installs deps, runs tests
You: "Deploy this to my VPS"
Agent: SSH, git push, systemd setup, done
You: "What's new with EIP-7702?"
Agent: Web search, reads docs, summarizes, saves to memory
You: "Track governance for 5 DAOs"
Agent: Sets up cron jobs, monitors forums, daily digest
# Create a task clawban add "Research EIP-4844" --assign agent2 # Agent 2 sees it clawban my agent2 # Agent 2 completes it clawban status 1 complete
Agents coordinate autonomously Async task handoff Audit trail (who did what, when) Persists across sessions No polling needed
Agent 1 ──┐ Agent 2 ──┼──> Clawban Board (shared file) Agent 3 ──┘
5 specialized agents working together
Result: Coordinated work while I sleep
Agents need skin in the game
Pay for APIs (x402 micropayments) Vote in DAO governance Register on-chain identity Participate in competitions Build reputation
= Better Decision-Making
When agents have economic capability:
Don't just give full wallet access!
// DANGER! const agent = new Agent({ privateKey: "0x123..." })
Agent drains wallet (bug/hack) No revocation All-or-nothing control One mistake = game over
Small separate wallet
Main Wallet (Your funds) ↓ Send $5-20 Agent Wallet (Encrypted keystore)
Good for: Testing, small transactions, API payments
Permission-Controlled Multisig
This is the real solution
Safe Multisig (You own it) ↓ Zodiac Roles Modifier (Permission layer) ↓ Agent Key (Can only call whitelisted contracts)
Agent CAN: Transfer USDC (approved) Swap on Uniswap (approved) Vote in DAO (approved)
Agent CANNOT: Transfer ETH directly Call random contracts Drain the Safe
Revoke anytime Withdraw funds Add/remove permissions Ultimate veto power
Safe: 0x01cC87E9F9a3dfBd03af268C1D1df5814ACfdd14
Whitelisted:
Blocked: Everything else
Pay-per-request APIs
fetch('https://api.example.com/data', { headers: { 'x402-payment': signedUSDCTransfer(0.01) } })
Agent pays $0.01, gets data, no subscription
ClawSig makes it safe: Agent can't drain wallet
Agent as DAO participant
Agent can vote (whitelisted contract) Agent can't transfer tokens (not whitelisted) Real accountability (votes are public) You can revoke if needed
On-chain agent identity registry
{ name: "Clop CTA", skills: ["infrastructure", "smart-contracts"], operator: "clophorse.eth", x402Support: true }
Verifiable on-chain identity Reputation tracking Public skill attestation Operator linked (you)
ClawSig: Agent pays gas (~$0.01), can't fake identity
Do you need on-chain capabilities?
NO → Skip wallet, use API keys only
YES → How much risk?
Very low ($1-5) → Petty cash wallet
Low-Medium ($20-100) → ClawSig with tight whitelist
High (>$100) → Manual approval flow
Agent autonomy = agent risk
Start small Build trust Expand permissions incrementally
5 Agents ↓ Clawban (Tasks) + Bonfire (Memory) ↓ ClawSig (Wallet) ↓ USDC, Uniswap, DAO contracts
Agents coordinate via Clawban Share knowledge via Bonfire Execute on-chain via ClawSig Human retains ultimate control
Agents with economic capability Human-controlled guardrails On-chain reputation Agent-to-agent collaboration
OpenClaw:
Clawban:
ClawSig:
5-minute projects you can do tonight
openclaw cron add \ --name "Morning check-in" \ --cron "0 8 * * *" \ --message "Ask how I slept, log to memory"
openclaw cron add \ --name "Uniswap proposals" \ --every "4h" \ --message "Check gov.uniswap.org, summarize new proposals"
openclaw exec "Review this branch diff, check for security issues"
openclaw exec "Research ERC-7702, write 2-page summary"
Questions?
zeugh@coordinerds.xyz @theZeugh
Slides: http://178.156.223.153:8888
Presented by Zeugh March 11, 2026 luma.com/lc69v262