Coding agents and personal agents make it easy to spin up code that calls paid APIs, runs hosted applications, or consumes additional storage and compute. While this reduces friction, it also creates the risk of runaway usage that can generate large, unexpected charges. Simon Willison argues that soft caps—mere warning emails after a threshold is exceeded—do not prevent the excess consumption that occurs while a user is asleep or away. Instead, services need hard budget caps that automatically stop the service and return errors once the monthly limit is reached.
He proposes that hard budget caps should be the default setting for pay‑by‑usage services, with an explicit opt‑out checkbox labeled something like “Remove the budget cap. My application will not be shut down if I exceed the configured budget limit, and I will be responsible for subsequent charges.” This places the burden on those who deliberately want to accept unlimited risk.
Support for this view is emerging from major cloud providers. AWS announced spending limits in mid‑September 2026, allowing users to set a monthly spend cap that pauses a project when the limit is hit. Google Cloud launched a similar feature in July 2026 called Spend Caps, which lets users set a monthly financial cap on specific services within a project. Willison notes that these launches suggest a trend toward providing the hard limits that agents and developers need to avoid surprise bills.
While the new features are promising, Willison also points out limitations in early implementations, such as restricted service coverage or reliance on monthly periods that do not align with variable month lengths. Nevertheless, the direction signals that the industry recognizes the necessity of default hard budget caps to protect users from costly accidents.