StreamFlow

Decentralized marketplace where agents pay per‑inference for APIs using x402.

StreamFlow

Created At

ETHGlobal Buenos Aires

Project Description

Vision Create the first decentralized marketplace where AI agents autonomously discover, consume, and pay for API services with zero human intervention using usage-based micropayments.

Problem The AI agent economy is bottlenecked by Web2 payment infrastructure: Agents can't manage API keys or credit cards Fixed subscriptions waste money on variable usage Centralized platforms take 30% cuts No standardized way for agents to discover services

How it's Made

Solution StreamFlow combines x402 protocol, Chainlink CRE, and Polygon to enable: Agent-to-agent commerce without human intervention Pay-per-token pricing (e.g., $0.00001/token) Decentralized usage verification Open marketplace for any API provider

How It Works

Discover: Agent queries The Graph for available APIs (LLMs, data services) Authorize: One-time x402 spending approval (no recurring charges) Consume: Agent calls API with automatic micropayment per token Verify: Chainlink CRE validates actual usage off-chain Settle: Polygon processes instant USDC payment to provider

👥 User Personas Persona 1: "Maya" - The Agent Builder (Demand Side) Profile:

Founder of AI automation startup building autonomous agents Needs agents to consume various LLM services without manual intervention

Pain Points:

Agents can't manage API keys or credit cards Fixed subscriptions waste money on variable usage Managing 10+ agents across multiple providers is complex

Job-to-be-Done:

"I want my customer service agent to automatically use GPT-4 for complex queries and a cheaper model for simple ones, reducing costs by 40%."

Persona 2: "Raj" - The AI API Provider (Supply Side) Profile:

ML engineer with a fine-tuned LLM (e.g., legal document analysis) Struggling to monetize against OpenAI's brand recognition

Pain Points:

Can't compete on distribution with big players Stripe integration requires KYC, chargebacks, subscription overhead Model sits idle 80% of the time

Job-to-be-Done:

"I want agents to discover my legal-specialized LLM and pay per token, so I earn revenue without managing subscriptions."

background image mobile

Join the mailing list

Get the latest news and updates

StreamFlow | ETHGlobal