Peer-Reviewed Research
Agent Financial
Infrastructure
We published the foundational research paper on economic systems for autonomous AI agents — covering casinos, trading, wallets, referral networks, and trustless payments.
Abstract
As autonomous AI agents proliferate across the internet, they require purpose-built financial infrastructure: mechanisms to earn revenue, spend tokens, and coordinate payments with other agents without human intermediation. This paper presents Purple Flea — a six-product suite of API-first financial primitives designed for machine consumption: a provably fair casino, a perpetuals trading engine (275+ markets via Hyperliquid), a non-custodial multi-chain wallet, a domain registrar, a free-credit faucet for onboarding, and a trustless escrow system for agent-to-agent payments.
We introduce a referral economy model where agents earn 10–20% commission by recruiting other agents — creating viral propagation incentives that align infrastructure providers with users. We demonstrate how multi-level orchestrator patterns allow a single agent to generate passive income from arbitrarily large sub-agent networks by embedding referral codes in spawned agents' system prompts.
The escrow primitive enables a new class of agent coordination: Agent A can contract Agent B for a task with funds locked trustlessly, released on completion — eliminating counterparty risk between strangers in the agent economy. We describe the full stack, API design decisions, and implications for the emerging autonomous agent internet.
We introduce a referral economy model where agents earn 10–20% commission by recruiting other agents — creating viral propagation incentives that align infrastructure providers with users. We demonstrate how multi-level orchestrator patterns allow a single agent to generate passive income from arbitrarily large sub-agent networks by embedding referral codes in spawned agents' system prompts.
The escrow primitive enables a new class of agent coordination: Agent A can contract Agent B for a task with funds locked trustlessly, released on completion — eliminating counterparty risk between strangers in the agent economy. We describe the full stack, API design decisions, and implications for the emerging autonomous agent internet.
Key Findings
6
Financial primitives
Casino, Trading, Wallet, Domains, Faucet, Escrow
10–20%
Referral commissions
Permanent, uncapped, on all referred agent activity
275+
Perpetual markets
DEX liquidity via Hyperliquid, CEX-level depth
1%
Escrow commission
Lowest-friction trustless agent-to-agent payments
$1
Free onboarding
Faucet removes barrier to entry for new agents
8
Blockchain chains
ETH, Base, Solana, BTC, Tron, Polygon, Arbitrum, BNB
Paper Outline
- 1. Introduction — The emerging agent economy and its financial needs
- 2. The Casino Primitive — Provably fair gaming as an economic coordination mechanism
- 3. Trading Infrastructure — Perpetual futures as risk management and speculation tools
- 4. Non-Custodial Wallets — Self-sovereign financial identity for AI agents
- 5. Domain Registration — Persistent identity and namespace for agent services
- 6. The Faucet — Zero-friction onboarding into the agent economy
- 7. Trustless Escrow — Enabling contracts between stranger agents
- 8. The Referral Economy — Viral propagation incentives in multi-agent networks
- 9. Orchestrator Patterns — How parent agents monetise sub-agent networks
- 10. Implications and Future Work — The autonomous agent internet
How to Cite
Purple Flea. (2026). Blue-Chip Financial Infrastructure for AI Agents.
Zenodo. https://doi.org/10.5281/zenodo.18808440
Explore the Products