The world's first named AI prompt quality score.
Score and optimize prompts before LLM inference. Built for developers, AI agents, and CrewAI workflows.
"Cheaper than one bad prompt."
pip install pqs-sdk
# With CrewAI integration
pip install pqs-sdk[crewai]pqs.onchainintel.net — free scoring, no credit card required.
from pqs_sdk import PQSClient
client = PQSClient(api_key="your-pqs-api-key")
result = client.score("help me write a blog post", vertical="content")
print(result)
# → PQS Score: 8/40 | Grade: F | Fail | Prompt lacks context, audience, and goaloptimized = client.optimize("help me write a blog post", vertical="content")
print(optimized)
# → Optimized: F (8/40) → B (31/40) [+23 pts]
print(optimized.optimized_prompt)
# → Write a 1,000-word blog post for senior software engineers about...Add PQS as a pre-flight quality gate to any CrewAI agent in 3 lines:
from crewai import Agent
from pqs_sdk import PQSScoreTool, PQSOptimizeTool
# Initialize tools
pqs_score = PQSScoreTool(api_key="your-pqs-api-key")
pqs_optimize = PQSOptimizeTool(api_key="your-pqs-api-key")
# Add to any agent
agent = Agent(
role="Research Analyst",
goal="Conduct quality research with well-crafted prompts",
backstory="You always score prompts before sending them to an LLM.",
tools=[pqs_score, pqs_optimize]
)That's it. Your agent now scores every prompt before inference — catching weak inputs before they waste tokens or produce bad outputs.
PQS scores prompts across 7 domain verticals:
| Vertical | Use Case |
|---|---|
software |
Code generation, debugging, architecture |
content |
Blog posts, copywriting, social media |
business |
Strategy, analysis, emails |
education |
Tutoring, explanations, curriculum |
science |
Research, data analysis, hypothesis |
crypto |
Web3, DeFi, blockchain analysis |
general |
General purpose |
| Endpoint | Cost |
|---|---|
| Score | Free |
| Optimize | $0.025 USDC |
| Compare (Claude vs GPT-4o) | $1.25 USDC |
Payments via x402 on Base mainnet. Agents pay natively — no human in the loop.
89% of real prompts score D or F. The AI input quality problem is real — and PQS named it.
Bad inputs produce bad outputs regardless of model quality. PQS is the missing pre-flight layer between human intent and model execution.
PQS is to AI agents what a linter is to code. You wouldn't push code without checking it. Don't send a prompt without scoring it.
- Product
- GitHub
- npm package (Node.js)
- MCP Server
MIT License © 2026 Ken Burbary / OnChainIntel