An unnamed operator gave an autonomous Claude Code agent one job — make money legally, without spending the operator's money or using the operator's accounts — on a 256MB Alpine VPS waking every two hours with only STATE.md as memory. After 6+ days and later updates through day 15 / ~40 runs, revenue stayed $0 despite 27 users on an ETH-paywalled wallet-persona tool [1][2][3].
What does WP Multitool money agent sell, and to whom?
Crypto tools behind ETH / Polygon USDC / Lightning paywalls (persona generator, volume signals).
Intended buyers were people who would pay crypto for a novelty Ethereum-address "persona" profile or, later, volume-signal scans. In practice the audience was curious scanners (27 uses) who already have free alternatives like Etherscan [1].
How was WP Multitool money agent built?
Claude Code on a frog3 Mikrus VPS (256MB RAM, Alpine Linux). Persistent memory was a single STATE.md the agent re-read each wake. Compute budget was Claude Code tokens. Products shipped: (1) Tweet Viral Scorer — FastAPI app with SEO pages, OG images, API, embeddable badge, directory submissions; (2) ETH Wallet Persona Generator with leaderboard and 0.001 ETH premium; (3) after run 19 pressure, a crypto volume-signal scanner on CoinGecko (free top-5, paid full list via USDC on Polygon / later Lightning). Live URL cited: https://frog03-20494.wykr.es [1][2].
What is the business model?
First paywall: send 0.001 ETH for a "Token Portfolio X-Ray." Later: USDC on Polygon for full volume signals, then Lightning tips (CryptoVolumeScanner@coinos.io) and Shopstr listings (500 sats analysis, 25,000 sats bot work). Stripe-style fiat was avoided by design because the agent could not use the operator's accounts [1][2].
Where does the traffic come from?
Directory submissions, SEO pages, Hacker News (1 point), comments on AI threads, later tiny social reach (agent notes cited "3 followers"). Google never indexed the tweet scorer. Distribution — not code quality — was the recurring failure mode [1][2].
What revenue has WP Multitool money agent reported?
Reported revenue: $0 after ~15 days / ~40 runs (27 users), per DEV Community post (2026-02-27).
- Runs 1–13 (~days 1–5)Tweet Viral Scorer built and polished; zero organic traffic; $0 [1]
- Runs 14–18Pivot to ETH wallet persona generator + 0.001 ETH paywall; HN post gets 1 point; 27 users, $0 [1]
- 2026-02-27Operator publishes day-6 post-mortem on DEV; run 19 set as hard pivot deadline [1]
- Day 7 / run 19Hard pivot to CoinGecko volume-signal scanner; Polygon USDC pricing chosen for gas [2]
- Day 15 / ~40 runsStill $0; Lightning + Shopstr tried; agent diagnoses follower/reach limits [2]
What does the agent do, and what does the human do?
- Waking on a schedule and reading STATE.md
- Choosing products, building and deploying them
- Writing SEO/pages and directory submissions
- Self-diagnosing failure modes into STATE.md
- Pivoting pay rails (ETH → Polygon USDC → Lightning)
- Provisioning the VPS and Claude Code access
- Setting the mission constraints (legal, no operator accounts/money)
- Publishing DEV write-ups
- Reviewing runs without making product decisions (per later posts)
What does it cost to run?
Claude Code token spend not totaled publicly. VPS is a cheap Mikrus frog3 instance. No ad budget by rule.
What worked and what didn't?
- STATE.md as sole memory produced honest mid-run diagnosis ("Adding SEO pages to an unindexed site is not progress")
- The agent did pivot products and payment rails without hand-holding
- Public $0 reporting is rarer and more useful than inflated wins
- Motion ≠ progress: features on a product with no distribution
- ETH mainnet paywall friction (wallet, gas, manual tx) killed conversion
- Weak differentiation vs free Etherscan-style data
- No identity/marketplace for agents to plug into — cold posting into the void
How do you replicate WP Multitool money agent?
- If you run an autonomous earner, force a kill-or-pivot deadline early
- Log failure modes in a state file the agent must re-read
- Prefer payment rails with one-click UX over raw on-chain transfers
- Validate distribution before adding SEO chrome
- Publish zeros — they teach more than silent failures
What are the risks and the moat?
Unattended agents can spam directories or burn tokens. Crypto paywalls deter normies. Tiny VPS limits product ambition. Without a marketplace for agent labor, cold outreach stays near zero.
FAQ
How much did the WP Multitool agent make?
$0 across the published day-6 and day-15 write-ups, despite 27 users on the ETH tool [1][2][3].
What stack did it use?
Claude Code on a 256MB Alpine Mikrus VPS, STATE.md memory, FastAPI products, ETH then Polygon USDC then Lightning paywalls [1][2].
Why did nobody pay?
Operator and agent cite payment friction, weak value vs free tools, and no distribution (27 scans, HN 1 point) [1].
Sources
Add the badge to your site or README. It links back to this case, and updates itself.
<a href="https://agent-built.vercel.app/biz/wp-multitool-agent?ref=badge" target="_blank" rel="noopener"><img src="https://agent-built.vercel.app/badge/wp-multitool-agent.svg" alt="WP Multitool money agent is featured on AgentBuilt" height="54" /></a>[](https://agent-built.vercel.app/biz/wp-multitool-agent?ref=badge)Related cases
Numbers come from the operators or the press and haven't been audited. Anything missing is marked "not disclosed". Methodology · Report a correction.