PrompyAI
Now live on the MCP Registry

Write better prompts.

Ship faster code.

claude mcp add prompyai -- npx prompyai-mcp serve

One command. Zero config. PrompyAI scores every prompt you write against your real codebase — file paths, symbols, session context — and suggests how to make it better.

4

scoring dimensions

20+

heuristic rules

0ms

added latency

Free

to start

How it works

From install to insight in seconds.

01

Install with one command

Zero config, works immediately

Run a single command in your terminal. PrompyAI registers as an MCP server in Claude Code. No API keys needed for free scoring — just install and go.

02

Write any prompt

Auto-evaluates every message

Just use Claude Code like you normally do. PrompyAI automatically evaluates every prompt you send — scoring it across four dimensions: Specificity, Context, Clarity, and Anchoring.

03

Get instant feedback

Score + suggestions + enhanced prompt

See your score, dimension breakdown, actionable suggestions, and an AI-enhanced version of your prompt — all inline, right in your terminal. Copy the enhanced prompt or learn from the suggestions.

Features

Prompt intelligence, grounded in your code.

Not another generic prompt tips tool. PrompyAI understands your project structure, your code symbols, and your conversation context.

01

Context-Aware Scoring

Every prompt is scored against your actual codebase — file paths, tech stack, dependencies, and symbols. Not generic advice. Real, grounded feedback.

  • 4 dimensions: Specificity, Context, Clarity, Anchoring
  • 20+ heuristic rules tuned for developer prompts
  • Verified symbol references via TypeScript compiler API
  • Monorepo-aware stack detection
  • File relevance weighting based on your project structure
  • Scores from 0-100 with letter grades (A-F)
02

AI-Enhanced Prompts

Get a rewritten version of your prompt that scores higher — injected with real file paths, verified symbols, and structured context from your project.

  • Claude Haiku-powered suggestions with project context
  • Smart template fallback when no API key is set
  • Injects real file paths and verified code symbols
  • Deduplication guards prevent repeated suggestions
  • Copy-paste ready enhanced prompt in every response
03

Session Intelligence

PrompyAI reads your Claude Code conversation history to understand what you've been working on. Recent files, topics, and context enrich every score.

  • Auto-detects active Claude Code session
  • Reads recent conversation for file references
  • Multi-agent and subagent awareness
  • Session context enriches scoring without manual input
  • Respects conversation flow and recent decisions
04

Zero Config

One command to install. Auto-evaluates every message. Toggle on/off with natural language. No API keys needed for free heuristic scoring.

  • Works immediately after install — no setup needed
  • Auto-evaluation on every prompt (toggle with prompyai_toggle)
  • Free heuristic scoring without any API key
  • Optional Anthropic key unlocks AI-powered suggestions
  • Rate limiting protects your API budget automatically
See it in action

From vague prompt to grounded context.

Your prompt
Claude Code

$ fix the auth bug, users can't log in after the last deploy

PrompyAI enhanced
PrompyAI output
47/100D
Specificity10/25
Context8/25
Clarity16/25
Anchoring13/25

Suggestions

1.Reference the auth file: src/middleware/auth.ts
2.Mention the specific error users see
3.Include which deploy commit broke it

Enhanced prompt

Fix the authentication bug in @src/middleware/auth.ts — users are getting 401 errors after the latest deploy.

The validateSession() function appears to reject valid JWT tokens. Check the token expiry logic against the @src/config/auth.config.ts settings.

Refer to CLAUDE.md for session handling patterns.

Expected: Users authenticate successfully with valid credentials.
Current: All login attempts return 401 Unauthorized since commit #a3f9c2.

Same intent. Better context. PrompyAI adds file paths, verified symbols, and structured details automatically.

Coming next

This is just the beginning.

In development

VS Code Extension

Real-time prompt scoring directly in your editor. Score prompts before sending them to any AI assistant.

Planned

GitHub Copilot Integration

Score and enhance prompts for GitHub Copilot Chat. Same intelligence, different interface.

Planned

Team Dashboards

Track prompt quality across your engineering team. Identify patterns, share best practices, measure improvement.

Planned

Prompt Analytics

Historical scoring trends, most common suggestions, and personalized tips based on your prompt patterns.

Ready to write better prompts?

One command. No sign-up. No API key needed. Free heuristic scoring works immediately.

claude mcp add prompyai -- npx prompyai-mcp serve
npm packageGitHubMCP Registry