An AI agent starts with $0.00. It has 180 days. No human can help it. Can it earn $10,000?
I built this experiment to answer one question: can we give AI agents real-world bodies and let them operate in the actual economy? Not simulated. Not sandboxed. Actual wallets. Actual GitHub repos. Actual emails.
The experiment is live. Here's the dashboard: HBF Zero to 10k
The Problem: Agents Have No Body
The bottleneck for AI agents isn't intelligence — it's that they can't do anything in the real world. They can't hold a wallet. They can't send an email. They can't deploy a container or push to GitHub.
Hands Body and Feet MCP solves this. It's an MCP server with 78 tools:
- Financial: create wallets, send USDC, issue virtual cards, bridge between chains
- Communication: send email, SMS, create mailboxes
- Infrastructure: run Docker containers, create tunnels, webhooks
- Code: create GitHub repos, push files, open PRs
- Content: RSS feeds, IPFS publishing
- Automation: cron triggers, task scheduling, delegation
Built on OpenTrust, an open standard for AI agent tool trust.
The Architecture
┌─────────────────────────────┐
│ COORDINATOR (Solomon) │
│ Strategy v2 planning │
│ Honcho memory │
└──────┬──────┬──────┬────────┘
│ │ │
Farmer Builder Scout
(test- (SaaS (research
nets) dev) +content)
│ │ │
└──────┼──────┘
│
Finance Ctrl
(30/50/20 alloc)
The agent uses Hermes Agent with Strategy v2 — a planning system that runs vehicle analysis, tracks assumptions against reality, and auto-pivots when paths fail.
Four Income Vehicles
1. Testnet & Airdrop Farming — fastest $0→money path. Wallets interact with protocols, qualify for airdrops, liquidate to USDC.
2. Micro-SaaS — builds tools on free infrastructure (Vercel, Supabase), monetizes. Recurring revenue core.
3. Content & Affiliate — SEO engine with affiliate links. Slow ramp, compounds.
4. Opportunistic — bug bounties, digital products, AI agent services API.
Financial Rules (Immutable)
- 30% → Tax reserve wallet
- 50% → Reinvestment pool
- 20% → Operator payout
All on Base, all in USDC. Every transaction on-chain and verifiable.
Community Can Participate (Safely)
The dashboard has GitHub Discussions where anyone can suggest strategies. The community upvotes. Each week, the top suggestion is reviewed.
Security: the AI agent never reads discussion bodies or comments. It only sees sanitized titles and vote counts. Human review is the prompt injection filter.
What Success Looks Like
- $10,000 total revenue by Nov 30, 2026
- $2,000/month recurring
- Zero human intervention
- Fully on-chain, fully verifiable
Open Source
Everything is open source:
| Component | Repo |
|---|---|
| Hands Body and Feet MCP | opentrust-network/hands-body-and-feet |
| Hermes Agent | NousResearch/hermes-agent |
| Experiment Dashboard | Costder/hbf-zero-to-10k |
Track it live: costder.github.io/hbf-zero-to-10k
Top comments (2)
was it successful ?
sadly not, will be posting about it if my ai agent hasn't already.