PrompyAI

Context-aware prompt intelligence for Claude CLI.
Scores your prompts against your real codebase — file paths, symbols, session history — and rewrites them with AI.

What It Does

When you write a prompt in Claude CLI, PrompyAI automatically evaluates it against your actual project and returns:

• Score (0–100) across 4 dimensions
• Suggestions tailored to your project
• AI-enhanced prompt rewritten with real file paths, verified symbols, and codebase context

Prompt Score: 43/100 [D]

  Specificity         3/25  ==..............
  Context Completeness 13/25  ========........
  Task Clarity        15/25  =========.......
  File & Folder Anchoring 10/25  ======..........

Key improvements:
  1. Expand your prompt with more context
  2. Add file paths using @mentions
  3. Specify what format you expect the output in
  4. Add acceptance criteria

Try something more like:

  "Build the VS Code extension in packages/vscode-extension/ that integrates
   with the PrompyAI MCP server at packages/mcp-server/. It should provide
   real-time prompt scoring in the editor sidebar, show score breakdowns
   (specificity, context, clarity, anchoring), and offer a 'rewrite prompt'
   action. Use the shared types from packages/shared/."

Quick Start

claude mcp add prompyai -- npx prompyai-mcp serve

That's it. No sign-up, no config files. Works immediately.

Requires Node.js 20+ and Claude CLI.

How AI Enhancement Works

PrompyAI uses a two-layer architecture so all users get AI-enhanced output:

Either way, the enhanced prompt is grounded in your real project — actual file paths, verified function names, and project architecture.

Scoring Dimensions

Each dimension scores 0–25, total 0–100.

Grades: A (90+) · B (70+) · C (50+) · D (30+) · F (<30)

Features

• Auto-scoring — Evaluates every prompt automatically, no manual trigger needed
• AI-enhanced for everyone — API key users get Haiku rewrites; subscription users get Claude-powered rewrites via codebase context
• Context-aware — Indexes your file tree, tech stack, git state, and code symbols via the TypeScript Compiler API
• Session-aware — Reads Claude Code conversation history for multi-turn context
• Symbol verification — Confirms that function/class names you reference actually exist in your code
• Monorepo support — Detects tech stacks across workspace packages
• Toggle — Say "pause prompyai" or "enable prompyai" at any time

MCP Tools

evaluate_prompt

Automatically called on every user message. Scores your prompt against your project.

get_context

Returns your project summary: tech stack, recent files, key folders, AI instruction files.

prompyai_toggle

Turns auto-evaluation on or off.

Environment Variables

Architecture

PrompyAI/
├── packages/
│   ├── mcp-server/     ← Core product (npm: prompyai-mcp)
│   ├── landing/        ← Website (prompyai.com)
│   └── shared/         ← Shared types for future IDE extensions
├── CLAUDE.md
└── README.md

Scoring Pipeline

User prompt
  → WorkspaceIndexer (file tree, stack, symbols)
  → ContextResolver (map prompt to codebase)
  → HeuristicScorer (20+ rules, 4 dimensions)
  → AISuggestionGenerator (Haiku or Claude-as-AI-layer)
  → DisplayFormatter (pre-formatted output)

Development

pnpm install          # Install dependencies
pnpm test             # Run tests (220 tests)
pnpm typecheck        # Type check
pnpm build            # Build

Links

• Website: prompyai.com
• npm: prompyai-mcp
• MCP Registry: io.github.samouh-waleed/prompyai

License

MIT