Buyer agent — decision log
Every tool call the procurement agent makes — discovery, ranking, payment, feedback — streamed live as it happens.
What you're looking at: a real Claude model, given a task in plain English (e.g. "buy market intel and leave feedback"), decides on its own which provider to use, spends real testnet money, and writes a review on-chain. Each card below is one tool call it chose to make — this is not a scripted replay.
connecting…
Spent so far: $0.000000
No decisions logged yet. Run
pnpm --filter buyer-agent start -- "risk brief on pool X" --budget 0.02 --live in another terminal to see it appear here in real time.