Free self-check
The Spend Guard self-check.
Five questions. Answer for the agents you run today, the ones that can spend money or call paid APIs. The result scores the four controls that decide whether a runaway agent is a bad hour or a five figure bill, and names the gaps in your answers. Nothing is stored when you run the check.
No gaps in your answers. The next step, if you want it checked against your real setup rather than your memory of it, is the free read-only audit.
The audit is read-only: an assessment of your agent spend paths and a signed evidence report of the gaps. It does not change or stop any spend. The gateway guard is the fix, offered after the audit.
Why these four controls
In the public runaway spend cases, the same controls are named missing each time: a spend ceiling, a kill switch, and an accountable owner. A runaway accounting agent documented by Mandiant burned about $50K in under an hour. A LangChain loop burned about $47K while dashboards watched. Read both cases: the Mandiant case and the LangChain case.