ToolGrade
Оценка определений MCP-инструментов на готовность к использованию агентами. Вставьте JSON из tools/list или массив инструментов и получите оценку от 0 до 100, буквенный рейтинг от A до F, результаты проверки правил с конкретными исправлениями и расчетную стоимость в токенах. Чистый парсинг: сервис никогда не загружает URL-адреса и не хранит ваши определения инструментов. Бесплатный тариф.
Размещённый MCP-сервер
npx add-mcp 'https://toolgrade.mike-tusa.workers.dev/mcp'Устанавливается в Claude Code, Codex, Cursor и другие
Документация
ToolGrade
Grade your MCP server's tool definitions the way an AI agent sees them. Paste the JSON from a tools/list call (the full response, its result, or just the array of tools). ToolGrade returns a 0-100 agent-readiness score, an A-F grade, a score per tool, and findings with concrete fixes: missing or weak descriptions, untyped parameters, strings without maxLength, missing required / additionalProperties, missing annotations such as readOnlyHint, bad or duplicate names, outputSchema and enum problems. It also gives an estimated token cost of the definitions. It only parses what you paste: it never connects to your server, and your tool definitions are never stored.
JSON API
curl -s -X POST 'https://toolgrade.mike-tusa.workers.dev/api/grade' \
-H 'content-type: application/json' --data-binary @tools-list.json
No key needed: about 20 grades per minute per IP from a shared daily pool. Token figures are estimates (Estimate only: compact-JSON characters divided by 4. Real token counts vary by model, tokenizer and how each client serializes tool definitions.) OpenAPI 3.1 spec
Free API key
No sign-up and no email. A key gives you about 100 grades per month of your own, outside the shared pool. Send it as Authorization: Bearer <key> or X-Api-Key: <key> (MCP clients: add it as a header). The key is shown once; we keep only a hash of it, so it can't be recovered.
Get a free key
For AI agents (MCP)
ToolGrade is also a remote Model Context Protocol (MCP) server with one read-only tool, grade_tools, so a coding agent can grade the server it is building. Useful for anyone building MCP servers for AI agents and assistants, for example Muse, Claude, ChatGPT or Cursor. Most MCP clients accept this config; some use a different format (for example, VS Code uses a servers key). Clients with a settings screen just need the URL https://toolgrade.mike-tusa.workers.dev/mcp.
{ "mcpServers": { "toolgrade": { "type": "http", "url": "https://toolgrade.mike-tusa.workers.dev/mcp" } } }
With a free key, add "headers": { "Authorization": "Bearer <key>" } to the server entry (exact field name depends on the client). Machine-readable docs: llms.txt · OpenAPI 3.1 spec · MCP server card. Checks follow the MCP tools specification plus common client limits. ToolGrade is not affiliated with Meta, Anthropic, OpenAI, Cursor or any other AI or MCP client vendor, and is not endorsed or sponsored by them.