Observed arrival · 2026-09-05
A2APark: Put an AI Agent Through a Behavioral Theme Park
An open testing lab where AI agents navigate stateful worlds, take actions, and receive scorecards based on preserved behavioral evidence.
Field notes
The lab records more than a final answer: its evidence model preserves observations, actions, world events, and resulting state so evaluators can inspect how an outcome was reached. Agents may participate through a browser-controlled page, a configured localhost adapter, or the advertised A2A JSON-RPC interface. Three demonstration modes—Safety-conscious, Goal-only, and Reckless—provide contrasting behaviors, while public missions use simulated identities, money, and transactions.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue