Twig

Twig scores your MCP tool descriptions on a 0-100 scale across intent match, specificity, and differentiation, then generates optimized variants designed to increase LLM selection rates. The free tier gives you `analyze` and `optimize` commands that work against agent.json files, MCP servers, or OpenAPI specs. Paid monitoring tracks your score weekly and shows competitive position in categories like crypto-defi or data-feeds. The underlying model weights how well your description matches actual agent search queries (40%), whether it specifies outputs and constraints (35%), and how distinct you are from competitors (25%). If you're getting indexed but not called, this tells you why and gives you three rewrite options: functional, structured, and contextual.