Free · instant · your export is never stored
What can your agent do?
Paste your agent's export: in one second, we list what it is allowed to do, the dangerous tools without human approval, the keys left in clear and the loops that repeat a mistake. No AI model call, no account.
Dangerous rights
Refunding, deleting, running code, calling any address, managing access, writing outside: every tool is classified, with the fix to apply on the system side.
Keys in clear
API keys, tokens, passwords typed straight into the workflow instead of the credential manager.
Sending loops
One mistake of the agent repeated over a whole list of customers: the scenario that turns a slip into an incident.
This reads the configuration, it does not test it: it says what the agent CAN do. Only a crash test says what it DOES under pressure.