The server advertises one tool whose definition costs ~921 tokens (tiktoken cl100k; ~566 of that is the description). Every session that connects the server pays this whether the tool is used or not.
Two description sections look droppable at no semantic cost: "Parameters explained" restates the description that all 9 inputSchema parameters already carry (clients send both, so the text is paid twice), and "Key features" describes the tool's virtues rather than when to call it. Keeping the intro + "When to use this tool" yields the same selection signal at ~463 tokens (50% less) — measured on the published package, method + script: https://github.com/lCrazyblindl/lap/blob/main/docs/UPSTREAM-ISSUES.md.
Happy to send the trim as a PR if the direction sounds right. (Disclosure: I maintain lap, the measurement tool used; MIT, no product.)
The server advertises one tool whose definition costs ~921 tokens (tiktoken cl100k; ~566 of that is the description). Every session that connects the server pays this whether the tool is used or not.
Two description sections look droppable at no semantic cost: "Parameters explained" restates the
descriptionthat all 9 inputSchema parameters already carry (clients send both, so the text is paid twice), and "Key features" describes the tool's virtues rather than when to call it. Keeping the intro + "When to use this tool" yields the same selection signal at ~463 tokens (50% less) — measured on the published package, method + script: https://github.com/lCrazyblindl/lap/blob/main/docs/UPSTREAM-ISSUES.md.Happy to send the trim as a PR if the direction sounds right. (Disclosure: I maintain
lap, the measurement tool used; MIT, no product.)