All 5 scenarios passed

An AI agent ran real shopping tasks through the MCP server and paid through PayPal Sandbox. Every payment was checked against PayPal and the store's database.

site

demo-store

agent

openai/gpt-oss-20b

store bugs

OFF

run

r_20261112_000637 · 12 Nov 2026, 05:39

scoring

money scenarios count double