What an MCP server's tool list costs: we counted the tokens of 34,374 of them
Описание и применение
Список инструментов MCP‑серверов может достигать сотен тысяч токенов, что приводит к высоким затратам на запросы. Кэширование и загрузка определений по требованию позволяют сократить расходы до 95 %.
Задача
Clients pay for sending large tool lists with every model request, inflating token usage and cost.
Результат
Typical MCP server tool lists average 1,272 tokens (median) but can reach over 500,000 tokens, causing significant cost per request. Strategies like prompt caching and on-demand loading can reduce costs by up to 95%.
Рабочий процесс
- Count tokens of each tool list using o200k_base tokenizer.
- Calculate tokens per server and per tool.
- Estimate cost per 1,000 requests for different model prices.
- Provide a counter tool to measure a specific server's list.
- Offer guidance on reducing token usage.
Доказательства
Under our survey of 12,915 MCP servers, a reader did the arithmetic we had not: 182,157 tool definitions over 7,656 servers is about 24 tools each, and a client resends every one of them with each model request.
The list rides along with every request, so its cost is tokens × price × requests.
Prompt caching. If the client keeps the tool list identical and early in the prompt, a cached read costs about a twentieth of the input price.
We put the counter on a free page: https://modelmarket.dev/tools/tool-list-cost
- Дата записи: 2026-10-19T18:41:05.760Z