Back to workshops
Free workshop90 minWatch recording
Tool schemas agents actually call
A free 90-minute live session on designing the tool surface an agent has to work through, and why most of them fail at the description field.
Watch the recordingWatch the session
We took three real tool definitions apart on the call and rewrote them live. The recording picks up from the first one.
What it covered
- Reading a failed tool call as an API review
- Naming, and why the description field does most of the work
- Splitting one tool that quietly does four things
- Error messages an agent can actually recover from
Who it is for
- Engineers whose agent picks the wrong tool more often than it should
- Anyone about to add a fifth parameter to a tool that already has four