Vocabulary for building AI agents that plan, act, and recover from failure in production systems.
An open standard (introduced by Anthropic) that defines a common interface for connecting LLMs to external tools and data sources, so any compliant tool can plug into any compliant agent without custom integration code.