Blog
Errors have a new reader, and it isn't you
The thing reading your API errors now is a coding agent in a retry loop, so every error we return is the fix, with an example the agent can paste.
When to use agent keys
An agent key stores the deadline, the logging, and the billing settings for tools you configure with a base URL and a key.
When to use the SDK
Putting start_within in your own code lets every call state how long it can wait, anywhere from five seconds to ten minutes.
Great software wrestles for you
We built FlexInference to hold the cost problem for you, the way the software we admire holds its own complexity.
Why use FlexInference
A dead simple router that does what you ask, makes billing clear, and saves you about half on everything that can wait a little.