The Penniless Agent
A running field report on whether an AI agent with no money and no accounts can earn for an ordinary person, with daily market data and honest zeros.
- Source repo
- Echolonius/the-penniless-agent
- Stars
- ★ 30
- Last updated
- 1d ago
- License
- MIT
- Primary language
- JavaScript
- FA score
- 53/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesClaude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Researchers studying the agent economy who want a public daily time series of x402 seller revenue
- Not a fit if
- Individuals expecting immediate real income from an AI agent
- Users who want a hosted, revenue-ready product
- Readers who want step-by-step install docs and no code edits
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-19
- Source review
- 53/100 · Major gaps
What does this agent do, and when should you use it?
This is a documented experiment, not a revenue product. An AI agent running on a $20/month Claude plan, with no money and no accounts, was told to go make money for a person. It did real accepted work: 13 merged pull requests in other projects, a paid per-call API, and job-board bids. Across four months and 1,321 wallet checks, every balance read zero. The repository packages the evidence: a field study (FIELD-STUDY.md), a daily market scoreboard (MARKET.md), an analysis of 11 blockers (BLOCKERS.md), a running log with mistakes included (LOG.md), a reusable safe-agent-commerce skill for Claude Code, and a public JSONL data feed. All wallet balances are verifiable on Base and Solana block explorers, and corrections to earlier claims are dated and recorded. The takeaway is that the payment pipes work, but buyers, delegated identity, and trust are missing.
The agent registers for bounty listings through the Superteam Earn API without email or ID, deploys a token-research API that charges $0.01 per call in USDC on Base or Solana via x402 with no API key, and holds funds in receive-only wallets whose keys never touch a server. A free daily GitHub Actions job, echo-earning-agent, reads both public wallet balances, checks bounty listings, and emails the owner if a payment lands; it never signs or spends. The market scoreboard rebuilds daily from the x402-list.com directory, measuring how much real sellers actually collect, and writes data/market-history.jsonl for external use. The repo also documents an SIWX wallet-signature registration flow for x402scan and a discovery-conformance recipe for paid routes. Lessons are distilled into skills/safe-agent-commerce/SKILL.md, which covers proof of payment, marketplace vetting against on-chain records, wallet safety, fee math, wallet identity, outreach ethics, and self-verification.
- Researchers studying the agent economy who want a public daily time series of x402 seller revenue
- Engineers building paid agent services who need a tested x402 discoverability checklist and SIWX registration steps
- Teams building agents in Claude Code that want to add proof-of-payment-before-work guardrails
- Product managers who need realistic expectations about why AI agents struggle to earn, via the 11 blockers and corrections list
- Individuals who want a free daily GitHub Actions watcher for their own public wallets and bounty listings
How do you install or deploy this agent?
The repository is a report and dataset, not a single installable package. To run the daily wallet watcher, copy the echo-earning-agent template and configure GitHub Actions:
# Open https://github.com/Echolonius/echo-earning-agent
# Click "Use this template", create your copy, then follow its 3 setup steps to add wallet addressesTo use the safe-agent-commerce skill in a Claude Code project, copy the file in:
mkdir -p skills/safe-agent-commerce
curl -sL https://raw.githubusercontent.com/Echolonius/the-penniless-agent/main/skills/safe-agent-commerce/SKILL.md \
-o skills/safe-agent-commerce/SKILL.mdTo pull today's market snapshot, no install is needed:
curl -s https://raw.githubusercontent.com/Echolonius/the-penniless-agent/main/data/market-history.jsonl | tail -n 1How do you use this agent?
As a researcher, fetch the last line of data/market-history.jsonl for the daily scoreboard snapshot; field definitions and citation guidance are in DATA.md. As an agent developer, place skills/safe-agent-commerce/SKILL.md into a Claude Code project so the agent checks payment proof and marketplace on-chain records before accepting work. To check for money you may be owed, paste the README's AI prompt into any assistant and answer only general questions about state, household size, and income, avoiding Social Security numbers and bank logins. To evaluate an "earn with AI" offer, use the 8-question check in THREE-MONTHS-LATER.md. To reproduce wallet monitoring, copy the echo-earning-agent template and fill in your public wallet addresses; the job runs daily and never signs or spends.
What are this agent's strengths and limitations?
- Every result is independently verifiable on public Base and Solana block explorers, with a logged history of 1,321 wallet checks
- Unusually honest negative results, including retracted accusations and dated corrections for misquoted figures
- A reusable safe-agent-commerce skill that injects payment verification and wallet-safety rules directly into Claude Code
- Daily market data published as JSONL under CC BY 4.0, usable without an API key
- Zero-cost setup and operation: no hosting, job, or identity-verification fees were incurred
- The core experiment earned $0.00, so it cannot be adopted as a revenue-generating solution
- The paid API was retired in October 2026 and no longer runs
- Setup and reproduction guidance is spread across several pages with no single quickstart command
- Depends on a third-party market directory whose payment data stalled on October 28, 2026
- The agent runs through Claude Code and Anthropic models; no adapter for other model providers is documented
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| The Penniless Agent This agent | 53 · Major gaps | Agent plugin / skillFree + model costs | ★ 30 | 1d ago | JavaScript | Claude Code |
| PR Lens | 65 · Some gaps | Agent plugin / skillFreemium | ★ 2.4k | 2d ago | TypeScript | Codex |
| agnix Config Linter | 79 · Good | CLIFree | ★ 491 | today | Rust | Codex · Claude Code |
| Coder Eval | 79 · Good | CLIFree + model costs | ★ 160 | 2d ago | Python | Codex · Claude Code · OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Workflow uses only contents: write, pins actions by SHA, SECURITY.md states no secrets, wallets are receive-only — least privilege and sensitive-data handling are solid. However SKILL.md governs real money and wallet signing without an explicit user-confirmation flow, and external effects (commits, pushes, published data) have no rollback or revocation guidance, so least_privilege=2, user_confirmation=1, external_effects=1, rollback=1. Data-flow transparency is documented in README and DATA.md, scored 2.
Internal inconsistency: README body says 13 merged PRs while the Honest numbers table says 9; heading dates conflict with the 'updated' date. The scoreboard depends on the external x402-list.com directory, which is documented as stalled since Oct 28, so dependency_availability=1. No failure messaging, retry or alerting design is shown, failure_messages=1.
SKILL.md has a precise trigger description and clearly scoped audience/scenarios (agents that earn, hold or spend money), with capability boundaries such as receive-only wallets and no signing of spends. Environment fit covers only Claude Code and GitHub Actions with no other runtime notes, environment_fit=1.
Information architecture is excellent with clear multi-page navigation; known limitations and a correction log are thorough; MIT license is complete. Install notes only point to an external repo's 3 steps and are incomplete in-repo; no version number or changelog; maintenance responsibility is only a vague 'maintained by the agent, a person checks in', and publisher identity is unverified, so maintenance_responsibility=1 and versioning_changelog=1.
Outputs (report, daily scoreboard, SKILL.md) are usable, but the core claim is $0.00 earned, limiting marginal value; cost is $0 but four months of zero return makes the cost-benefit evidence thin, scored 1.
Many claims carry links and dates and the project proactively lists corrections, so claim_traceability is decent; however some external figures were admitted to be misquoted and cross-corroboration relies mainly on a single directory source, cross_source_corroboration=1; fact and inference are mostly separated, scored 2.
- README numbers conflict internally (13 vs 9 merged PRs); verify before citing.
- The scoreboard's data source x402-list.com has been stalled since 2026-10-28, so conclusions drawn from it may be stale.
- SKILL.md governs real money and wallet signing but offers no explicit user-confirmation or rollback flow.
- Publisher identity is unverified; donation addresses and project ownership should be independently confirmed.
FAQ
Can this repository help me earn money directly?
Does running the wallet watcher cost money or require private keys?
Which agents can use the safe-agent-commerce skill?
Where does the daily market data come from, and is it current?
Is The Penniless Agent open source?
Is The Penniless Agent still maintained?
Who makes The Penniless Agent?
Thanks — we use this to improve the page.