AI导哈

迭代生成判别工作流MCP服务

这是一个MCP服务器服务，通过生成器与判别器的循环迭代自动优化内容质量，适用于需要多轮改进和质量阈值控制的AI工作流开发场景。

  • by T7NCKE
  • TypeScript
  • 本地部署
  • 更新于
在 GitHub 查看 tincke10/Barto-MCP

使用说明

MCP Ralph Workflow

MCP Server for iterative Generator-Discriminator workflows (Ralph Loop)

A Model Context Protocol (MCP) server that implements an iterative improvement loop where a Generator creates content and a Discriminator evaluates it against specified criteria until quality thresholds are met.

Features

  • Iterative Refinement: Automatic improvement loop with configurable thresholds
  • Multi-Provider Support: Works with Anthropic (Claude) and OpenAI models
  • Async Execution: Support for both synchronous and asynchronous workflow execution
  • Cost Control: Built-in cost estimation and limits
  • Observability: Structured logging, metrics, and tracing support
  • Security: Input sanitization and prompt injection protection

Quick Start

Prerequisites

  • Node.js 20+
  • pnpm
  • Redis (for state management)
  • PostgreSQL (for history, optional)

Installation

# Clone the repository
git clone <repository-url>
cd mcp-ralph-workflow

# Install dependencies
pnpm install

# Copy environment file
cp .env.example .env

# Edit .env with your API keys
# At minimum, set ANTHROPIC_API_KEY

Development

# Run in development mode with hot reload
pnpm dev

# Run tests
pnpm test

# Run tests with coverage
pnpm test:coverage

# Type check
pnpm typecheck

# Lint
pnpm lint

# Format code
pnpm format

Build

# Build for production
pnpm build

# Start production server
pnpm start

MCP Tools

ping

Test the server connection.

{
  "name": "ping",
  "arguments": {
    "message": "optional message"
  }
}

run_workflow

Execute an iterative generator-discriminator workflow.

{
  "name": "run_workflow",
  "arguments": {
    "task": "Write a product description for a smartphone",
    "criteria": [
      "Clear and concise",
      "Highlights key features",
      "Professional tone",
      "Under 200 words"
    ],
    "maxIterations": 10,
    "scoreThreshold": 0.85,
    "mode": "sync"
  }
}

get_status

Check the status of a workflow.

{
  "name": "get_status",
  "arguments": {
    "workflowId": "uuid-here",
    "includeHistory": true
  }
}

cancel_workflow

Cancel a running workflow.

{
  "name": "cancel_workflow",
  "arguments": {
    "workflowId": "uuid-here",
    "reason": "No longer needed"
  }
}

list_workflows

List workflows with filtering.

{
  "name": "list_workflows",
  "arguments": {
    "status": "running",
    "limit": 20
  }
}

Configuration

Environment Variables

Variable Description Default
ANTHROPIC_API_KEY Anthropic API key Required
OPENAI_API_KEY OpenAI API key Optional
DEFAULT_LLM_PROVIDER Default provider anthropic
REDIS_URL Redis connection URL redis://localhost:6379
DATABASE_URL PostgreSQL connection URL Optional
MAX_ITERATIONS_LIMIT Max iterations allowed 50
DEFAULT_SCORE_THRESHOLD Default quality threshold 0.85
MAX_COST_PER_WORKFLOW_USD Max cost per workflow 1.00
LOG_LEVEL Logging level info

MCP Inspector (Testing & Development)

Use the MCP Inspector to test the server connection and available tools:

# Build first (required)
npm run build

# Run with MCP Inspector
npm run inspector

This will open a web interface where you can:

  • View all available tools (ping, run_workflow, get_status, etc.)
  • Test tool calls interactively
  • Inspect request/response payloads
  • Debug connection issues

Manual Inspector Usage

# Using npx directly
npx @modelcontextprotocol/inspector node dist/index.js

# With environment variables
ANTHROPIC_API_KEY=your-key npx @modelcontextprotocol/inspector node dist/index.js

Claude Desktop Integration

Add to your Claude Desktop config:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "ralph-workflow": {
      "command": "node",
      "args": ["/path/to/mcp-ralph-workflow/dist/index.js"],
      "env": {
        "ANTHROPIC_API_KEY": "your-api-key"
      }
    }
  }
}

Verifying Connection

  1. Restart Claude Desktop after updating the config
  2. Look for the hammer icon (🔨) in the chat interface
  3. Click it to see available tools from ralph-workflow
  4. Test with: "Use the ping tool to test the connection"

Project Structure

src/
├── index.ts              # Entry point
├── server.ts             # MCP server setup
├── tools/                # MCP tools
├── core/                 # Business logic
│   ├── orchestrator/     # Workflow orchestration
│   └── runners/          # Generator/Discriminator
├── infrastructure/       # External services
│   ├── llm/             # LLM providers
│   ├── persistence/     # Redis/PostgreSQL
│   └── queue/           # BullMQ
├── shared/              # Shared utilities
│   ├── errors/          # Error types
│   ├── types/           # Type definitions
│   └── utils/           # Utilities
├── schemas/             # Zod schemas
└── config/              # Configuration

Testing

# Run all tests
pnpm test

# Run specific test file
pnpm test tests/unit/tools/ping.test.ts

# Run with coverage
pnpm test:coverage

# Watch mode
pnpm test -- --watch

License

MIT

相关 MCP