PQS MCP Server

The world's first named AI prompt quality score — as an MCP server.

Score, optimize, and compare LLM prompts before they hit any model. Built on PEEM, RAGAS, G-Eval, and MT-Bench frameworks.

Install

Claude Desktop

Add to your config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "pqs": {
      "command": "npx",
      "args": ["pqs-mcp-server"]
    }
  }
}

Smithery

smithery mcp add onchaintel/pqs

Tools

score_prompt (Free — no API key needed)

Score any prompt before it hits any model. Returns grade A-F, score out of 40, and percentile.

Example output:

{
  "pqs_version": "1.0",
  "prompt": "analyze this wallet",
  "vertical": "crypto",
  "score": 8,
  "out_of": 40,
  "grade": "D",
  "upgrade": "Get full dimension breakdown at /api/score for $0.025 USDC via x402",
  "powered_by": "PQS — pqs.onchainintel.net"
}

optimize_prompt ($0.025 USDC via x402)

Score AND optimize any prompt. Returns full 8-dimension breakdown + optimized version.

Requires: PQS API key (get one free at pqs.onchainintel.net)

compare_models ($1.25 USDC via x402)

Compare Claude vs GPT-4o on the same prompt. Judged by a third model. Returns winner, scores, and recommendation.

Requires: PQS API key (get one free at pqs.onchainintel.net)

Verticals

Specify the domain context for more accurate scoring:

• software — Software engineering, code, debugging
• content — Content creation, copywriting, social media
• business — Business analysis, finance, strategy
• education — Education, research, academic writing
• science — Scientific research, data analysis
• crypto — Crypto trading, DeFi, onchain analysis
• general — General purpose (default)

Quality Gate Pattern

Use PQS as a pre-inference quality gate:

const score = await fetch("https://pqs.onchainintel.net/api/score/free", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ prompt: userPrompt, vertical: "software" })
});
const { score: pqsScore } = await score.json();
if (pqsScore < 28) throw new Error("Prompt quality too low — improve and retry");

Grade D or below (< 28/40) means the prompt will waste inference spend.

Built by

John / OnChainIntel — @OnChainAIIntel
pqs.onchainintel.net