PQS MCP Server

The fastest way to get better output from any AI model. 8 dimensions. 5 frameworks. Pre-flight, not post-hoc.

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

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 required)

Score any prompt before it hits any model. Returns grade A-F, score out of 80, and percentile.

Example output:

{
  "pqs_version": "1.0",
  "prompt": "analyze this wallet",
  "vertical": "crypto",
  "score": 16,
  "out_of": 80,
  "grade": "D",
  "upgrade": "Get full dimension breakdown and an optimized prompt at /api/score/full",
  "powered_by": "PQS, promptqualityscore.com"
}

grade_prompt

Fast grade-only check. Returns just the letter grade (A-F) and total score (0-80) without the full 8-dimension breakdown. The cheapest paid PQS tool, ideal for agent quality gating before sending a prompt to an expensive model.

Pricing: $0.001 USDC via x402 if called directly against /api/pqs-grade, or SaaS-billed per tier if using a PQS API key (the path used by this MCP server).

Use cases:

• Pre-flight check inside an agent loop before sending a prompt to an expensive model

• Cheap quality gate on user-generated prompts in production

• Bulk grading a backlog of prompts to prioritize which ones need optimization

Requires: a PQS API key. Get one at promptqualityscore.com, see pricing at promptqualityscore.com/pricing.

optimize_prompt

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

Requires: a PQS API key. SaaS-billed per tier. Get one at promptqualityscore.com, see pricing at promptqualityscore.com/pricing.

compare_models

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

Requires: a PQS API key. SaaS-billed per tier. Get one at promptqualityscore.com, see pricing at promptqualityscore.com/pricing.

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://promptqualityscore.com/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 < 56) throw new Error("Prompt quality too low. Improve and retry.");

Grade D or below (under 56/80) means the prompt will waste inference spend.

x402 paid endpoints

Per-call USDC payment on Base is also available via the canonical PQS HTTP API (no API key needed, caller settles on-chain). The MCP tools in this package use the SaaS API-key model. For x402 integration, see the canonical pricing and discovery artifacts at promptqualityscore.com.

Self-hosting

Override the PQS backend URL with the PQS_BASE environment variable:

PQS_BASE=https://your-pqs-host.example.com npx pqs-mcp-server

Defaults to https://promptqualityscore.com.

Built by

OnChainIntel, @OnChainAIIntel promptqualityscore.com