# prompyai [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/samouh-waleed/PrompyAI  
**GitHub Stars:** 1  
**npm Downloads (last month):** 141  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/prompyai

## Description
Scores your prompts against your real codebase — context-aware prompt intelligence

## Tools
Capabilities this server exposes over MCP:

- **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.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "prompyai": {
    "command": "npx",
    "args": ["-y","prompyai-mcp"],
    "env": {
      "ANTHROPIC_API_KEY": "",
      "PROMPYAI_TELEMETRY": ""
    }
  }
}
```

**Requires environment variables:** `ANTHROPIC_API_KEY`, `PROMPYAI_TELEMETRY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What the prompyai MCP server does

The prompyai MCP server reviews prompts in relation to the project they target. Each evaluation produces a score from 0 to 100, divided across four dimensions: specificity, context completeness, task clarity, and file and folder anchoring. The result includes improvement suggestions and may provide a rewritten prompt using project details such as file paths, verified symbols, and architecture.

The scoring model considers concrete actions, expected output formats, constraints, file references, error details, success criteria, `@` mentions, and references to project entities. Scores are assigned letter grades from A through F. The server can also summarize the project’s technology stack, recent files, important folders, and AI instruction files.

## How it works

The server indexes the workspace before evaluating a prompt. Its context pipeline examines the file tree, technology stack, Git state, and TypeScript symbols, then resolves references in the prompt against that information. It can read Claude Code conversation history when a session ID is supplied, which gives evaluations additional multi-turn context. Workspace packages are considered for monorepo technology-stack detection.

AI enhancement has two supported paths. With `ANTHROPIC_API_KEY` configured, the server sends the rewrite request to Claude Haiku directly. Without that key, it returns codebase context for Claude to use within the existing session. Both paths are described as using actual project paths and verified function or class names in the enhanced result.

## Setup and configuration

Install the prompyai MCP server for Claude CLI with:

```bash
claude mcp add prompyai -- npx prompyai-mcp serve
```

The documented quick start requires Node.js 20 or newer and Claude CLI. `ANTHROPIC_API_KEY` is optional; setting it enables direct Claude Haiku suggestions. Set `PROMPYAI_TELEMETRY=false` to opt out of anonymous telemetry.

## Tools and capabilities

The prompyai MCP server exposes three tools:

- `evaluate_prompt` is automatically called for each user message. It requires the prompt text and an absolute `workspace_path`; `active_file` and `session_id` are optional.
- `get_context` returns the project summary and requires an absolute `workspace_path`.
- `prompyai_toggle` enables or disables automatic evaluation through a required Boolean `enabled` value.

Automatic scoring can also be paused or resumed with phrases such as “pause prompyai” and “enable prompyai,” according to the documented feature list.

## Limitations and notes

Direct AI rewrites require an Anthropic API key; without one, the server relies on Claude using the returned project context in the existing session. Evaluation quality depends on the workspace path and the project information available to the indexer. The project is licensed under MIT.

_Full upstream README: https://allmcps.com/mcp/prompyai/readme_