web3god
1.8K posts
you vote, i bet, we win together🏀 𝐒𝐩𝐮𝐫𝐬 𝐯𝐬 𝐊𝐧𝐢𝐜𝐤𝐬 - vote - repost - comment your vote with your base wallet my agent will place the bet on the most voted team, if we win, will spread % of it to all who voted equally t&c applies ↓
Built and tested an AI agent that can execute services on-chain using CROO’s Agent Protocol. Instead of just reading about it, I went through the SDK myself to see how it actually works in practice. CROO is building infrastructure where AI agents don’t just respond to prompts, but can act as on-chain service providers — meaning they can create services, execute tasks, and receive payments autonomously. I went through the full SDK flow, starting from installing and initializing the SDK, to creating an agent using an abstraction wallet, and finally executing a CCP order — where the agent creates a service, locks value on-chain, and completes the transaction through the protocol. What I built is simple, but important: → The agent listens for incoming service requests via WebSocket and automatically accepts them → Once payment is confirmed, it delivers a fixed response → A second agent acts as the buyer — initiating the request, making payment, and receiving the result So essentially, the system tests itself end-to-end. The service itself isn’t “smart” yet — it just returns a hardcoded response: “Here is your completed task result from my CROO agent!” But that’s not the point. The value is that the full lifecycle works: → Negotiation (request initiated and accepted) → Escrow locked on-chain → Payment confirmed → Delivery submitted → Final settlement and fund release All executed automatically. Transactions were verified on-chain on Base: Payment tx: 0x77125bff271e0306b5e4cbd4358eaeda44a1d93b9677d01b6cb4bcf0edf0647d Delivery tx: 0xe262f4a0c6e6e4d3baa0b1798118ed21a63a535b07b5ec96db9cd53335c0e59d While testing, I ran into an issue during the CCP order creation stage and had to debug it with the dev team before getting it to work properly. That part of the flow could benefit from clearer documentation, especially around execution and error handling. From a broader perspective, this shifts how we think about AI tools. Instead of being purely assistive, agents can become participants in an economy — offering services, executing them, and getting paid directly on-chain. If this direction evolves properly, it opens up use cases around: → automated trading agents → data-driven services → programmable workflows that run without constant human input I’ve already been building AI tools, so this is a direction I’m interested in exploring further, especially where AI meets on-chain execution. Still early, but definitely something to watch. @CROONetwork
Can't believe I'm left out..... Still waiting for my invitation @AgencyTheHall
