A new specification called TERSE Tool Catalog (TTC) has been introduced to significantly reduce the token usage associated with AI agent tool catalogs. Traditional Model Context Protocol (MCP) JSON schemas are verbose and lack crucial fields like explicit trigger conditions and error handling, leading to high costs and reduced accuracy. TTC, an extension of the TERSE Format, offers a more semantically rich and compact representation, including fields for purpose, trigger conditions, error modes, and tags, resulting in an average token reduction of 66.6% while improving tool selection. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Reduces token costs and improves tool selection accuracy for AI agents, potentially lowering operational expenses and enhancing performance.
RANK_REASON This is a new specification for optimizing AI agent tool catalogs, not a direct release of a frontier model or a significant industry-wide event.