{
  "id": 2761261,
  "title": "I Benchmarked 10 MCP Servers — One of Them Burns 47K Tokens Just to Say Hello",
  "url": "https://urgent.news/2026/08/23/i-benchmarked-10-mcp-servers-one-of-them-burns-47k-tokens-just-to-say",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-08-23T09:06:08.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/mcptokensaver/i-benchmarked-10-mcp-servers-one-of-them-burns-47k-tokens-just-to-say-hello-7he"
  },
  "original_language": "en",
  "account": "Testing 10 popular MCP servers revealed one server consumed 47,293 tokens just to list files in a user's Google Drive. After benchmarking, it became clear that MCP servers impose significant token overhead before any user queries. The main drivers of this overhead were JSON schema definitions, with 42% of tokens used for inputSchema properties and type definitions. Google Drive's MCP server had the most tools, each with deep nested schemas. The token cost for connecting 10 servers at $3 per million input tokens would amount to $0.34 - even exceeding the cost of a Claude Pro subscription. By building mcptoon, a CLI proxy that caches schemas and strips result wrapping, users can reduce token usage by up to 97%, saving significant costs per conversation.",
  "summary": "I Benchmarked 10 MCP Servers — One of Them Burns 47K Tokens Just to Say Hello 10 popular MCP servers. 847 tools total. 312K tokens of JSON schemas. One server alone wastes more tokens than a full GPT-3 conversation. Here are the results. What I did I installed the 10 most popular MCP servers from the official registry. Connected each one to a token counter. Measured exactly how many tokens get…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}