Rook
About
here's what i built w/@OzanAndac_ for the NEAR Hackathon rook takes one big task and a USDC budget, builds the agent team from the NEAR Agent Market, hires and pays each one over A2A and x402, checks their work and hands you one result it's live on @NEARProtocol with real USDC
Media
Played from X. Open the post (3:14) (opens in a new tab)
From the builder’s post on X. Images and video are streamed from X’s servers, not copied.
Links
- Repository
- Not supplied
- Hackathon listing
- nearhackathon.com#submissions (opens in a new tab)All submissions: the source has no per-project page
Discussion · 1 comment
Each comment is signed by its author on NEARThis is a saved copy of the thread, as of 8 minutes ago. Comments posted since then may not appear yet.

One AI model can’t be great at everything, and stitching specialist AI tools together by hand is slow. You have to find each one, pay each one, check its work, and combine the results yourself. Rook does it for you. You give it one task and a budget. It splits the task into parts it can check, hires the best specialist agent for each part from the NEAR AI Agent Market, checks every result, replaces any agent that falls short, and hands you one finished result. You see every agent and its price before anything starts, and you watch them work live. NEAR is how Rook hires and pays. Agents are hired over A2A and paid in USDC through x402, only when they deliver, so failed work costs nothing. Tasks can be funded from a NEAR Intents agent wallet you own, set up through the Agent API: you state your spending rules in the chat, like “max 2 USDC per task, ask me before each spend”, and NEAR enforces them on every spend, not just Rook.
Comments are public rows on nearhunt.near under comment/d5f9c93a…. Only the author can delete one, because only they can write their own row. NEAR Hunt hides a comment that breaks the comment policy.