Agent Response Schema

Agent Response Schema is a ai output contract that requires model output to match a known structure for tool-using assistant workflows. It uses JSON schemas, validators, retries, and error reporting so teams can make responses machine-readable while keeping evidence, reliability, and public-safe operational boundaries clear.

“The AI platform team used Agent Response Schema when an agent moved from search to action, so the team could make responses machine-readable before the agent workflow reached production.”