Jezweb MCP Core v3.0.1 - Adaptable Multi-Provider Architecture

A production-ready Model Context Protocol (MCP) server featuring an adaptable, provider-agnostic architecture that supports multiple LLM providers through a unified interface. Built with a "Shared Core with Thin Adapters" architecture for maximum flexibility and simplicity.

🌟 Universal MCP Server - Three Ways to Connect

Choose the deployment option that best fits your needs:

🚀 Option 1: Cloudflare Workers (Production Ready - v3.0 Unified Architecture)

Production URL: https://openai-assistants-mcp.jezweb.ai/mcp/{api-key}

• ✅ Adaptable Architecture - Support for multiple LLM providers (OpenAI, Claude, etc.)
• ✅ Simple Configuration - Environment-first configuration, no complex setup
• ✅ Lightweight & Fast - Sub-100ms response times with global edge distribution
• ✅ Zero Dependencies - No local setup required
• ✅ LIVE & OPERATIONAL - v3.0 unified architecture deployed and tested

📦 Option 2: NPM Package (Local Stdio - v3.0 Deployment Adapter)

Package: jezweb-mcp-core@3.0.1

• ✅ Provider-Agnostic - Unified core with deployment-specific adapter
• ✅ Simple Configuration - Environment variables and sensible defaults
• ✅ Direct stdio transport - No proxy required
• ✅ Local execution - Full control over environment
• ✅ 100% Backward Compatible - Seamless upgrade from OpenAI-specific versions

🔧 Option 3: Local Development Server

Local Build: Clone and run locally

• ✅ Full source code access
• ✅ Customizable implementation
• ✅ Development and testing
• ✅ Private deployment options

✨ Key Features - Jezweb MCP Core v3.0

🏗️ Adaptable Multi-Provider Architecture

• Provider-Agnostic Design - Support for OpenAI, Anthropic Claude, Google, and more
• Extensible Provider System - Easy to add new LLM providers
• Unified Interface - Same tools and resources across all providers
• Smart Provider Selection - Automatic fallback and load balancing
• Simple Configuration - Environment-first setup with sensible defaults

🚀 Core Capabilities

• Complete Assistant API Coverage - All 22 tools for full assistant, thread, message, and run management
• Universal Deployment - Three deployment options with identical functionality
• Production Ready - Deployed on Cloudflare Workers with modern architecture
• Lightweight - Minimal dependencies and fast execution
• Type Safe - Full TypeScript implementation with comprehensive type definitions

🎯 Enhanced User Experience

• Enhanced Tool Descriptions - Workflow-oriented descriptions with practical examples
• MCP Resources - 9 comprehensive resources including templates, workflows, and documentation
• Improved Validation - Detailed error messages with examples and suggestions
• Tool Annotations - Proper MCP annotations for better client understanding
• Assistant Templates - Pre-configured templates for common use cases

🔧 Technical Excellence

• Secure Authentication - URL-based API key authentication (Workers) or environment variables (NPM)
• Advanced Error Handling - Context-aware error messages with actionable guidance
• CORS Support - Ready for web-based MCP clients
• Real-time Operations - Support for streaming and real-time assistant interactions
• Comprehensive Testing - Built-in test suites for both deployment options

📊 Architecture Overview

Provider System

Jezweb MCP Core uses a sophisticated provider registry system that abstracts away provider-specific details:

// Multiple providers supported
const providers = {
  openai: { /* OpenAI configuration */ },
  anthropic: { /* Claude configuration */ },
  google: { /* Gemini configuration */ }
};

// Automatic provider selection
const provider = registry.selectProvider({
  strategy: 'capability-based',
  requiredCapabilities: ['assistants', 'threads']
});

Simple Configuration

Environment-first configuration with sensible defaults:

# Cloudflare Workers - via Wrangler secrets
wrangler secret put OPENAI_API_KEY
wrangler secret put ANTHROPIC_API_KEY

# NPM Package - via environment variables
export OPENAI_API_KEY="your-key-here"
export ANTHROPIC_API_KEY="your-key-here"

Unified Architecture

shared/                    # Unified shared core (single source of truth)
├── core/                  # Core business logic and handlers
├── services/              # Provider registry and LLM service abstraction
│   ├── llm-service.ts     # Generic LLM provider interface
│   ├── provider-registry.ts # Provider management and selection
│   └── providers/         # Individual provider implementations
└── types/                 # Unified type definitions

src/                       # Cloudflare Workers deployment
├── worker.ts              # Cloudflare Workers entry point
└── mcp-handler.ts         # Worker-specific MCP handler

npm-package/               # NPM package deployment
├── src/                   # NPM-specific implementation
└── universal-mcp-server.cjs # NPM package entry point

🚀 Quick Start - Choose Your Installation Method

Prerequisites

• API key for your chosen LLM provider (OpenAI, Anthropic, etc.)
• Node.js 18+ (for NPM package or local development)
• MCP client (Claude Desktop, Roo, or other MCP-compatible client)

🔑 Getting Started with LLM Providers

OpenAI Setup

• Visit the OpenAI API Keys page
• Create a new API key
• Monitor usage at OpenAI Dashboard

Anthropic Claude Setup

• Visit the Anthropic Console
• Create an API key
• Review Claude API documentation

📦 Option 1: NPM Package (Recommended for Most Users)

Installation

# Option A: Use directly with npx (recommended for latest fixes)
npx jezweb-mcp-core@latest

# Option B: Install globally
npm install -g jezweb-mcp-core@latest

# Option C: Install locally in your project
npm install jezweb-mcp-core@latest

Claude Desktop Configuration

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "jezweb-mcp-core": {
      "command": "npx",
      "args": ["jezweb-mcp-core@latest"],
      "env": {
        "OPENAI_API_KEY": "your-openai-api-key-here",
        "ANTHROPIC_API_KEY": "your-anthropic-api-key-here"
      }
    }
  }
}

Roo Configuration

Add to your Roo configuration file:

{
  "mcpServers": {
    "jezweb-mcp-core": {
      "command": "npx",
      "args": ["jezweb-mcp-core@latest"],
      "env": {
        "OPENAI_API_KEY": "your-openai-api-key-here",
        "ANTHROPIC_API_KEY": "your-anthropic-api-key-here"
      },
      "alwaysAllow": [
        "assistant-create",
        "assistant-list",
        "assistant-get",
        "assistant-update",
        "assistant-delete",
        "thread-create",
        "thread-get",
        "thread-update",
        "thread-delete",
        "message-create",
        "message-list",
        "message-get",
        "message-update",
        "message-delete",
        "run-create",
        "run-list",
        "run-get",
        "run-update",
        "run-cancel",
        "run-submit-tool-outputs",
        "run-step-list",
        "run-step-get"
      ]
    }
  }
}

☁️ Option 2: Cloudflare Workers (Zero Setup)

Claude Desktop Configuration

1. Install the MCP proxy:

npm install -g mcp-proxy

2. Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "jezweb-mcp-core": {
      "command": "npx",
      "args": [
        "mcp-proxy",
        "https://openai-assistants-mcp.jezweb.ai/mcp/YOUR_OPENAI_API_KEY_HERE"
      ]
    }
  }
}

🔧 Option 3: Local Development Server

Setup

1. Clone the repository:

git clone https://github.com/jezweb/openai-assistants-mcp.git
cd openai-assistants-mcp

2. Install dependencies:

npm install

3. Set up environment variables:

# Add your API keys to wrangler.toml or use wrangler secrets
wrangler secret put OPENAI_API_KEY
wrangler secret put ANTHROPIC_API_KEY

4. Start development server:

npm run dev

🛠️ Available Tools

Assistant Management

1. assistant-create - Create a new assistant with instructions and tools
2. assistant-list - List all assistants with pagination and sorting
3. assistant-get - Get detailed information about a specific assistant
4. assistant-update - Update assistant instructions, tools, or metadata
5. assistant-delete - Delete an assistant permanently

Thread Management

6. thread-create - Create a new conversation thread
7. thread-get - Get thread details and metadata
8. thread-update - Update thread metadata
9. thread-delete - Delete a thread permanently

Message Management

10. message-create - Add a message to a thread
11. message-list - List messages in a thread with pagination
12. message-get - Get details of a specific message
13. message-update - Update message metadata
14. message-delete - Delete a message from a thread

Run Management

15. run-create - Start a new assistant run on a thread
16. run-list - List runs for a thread with filtering
17. run-get - Get run details and status
18. run-update - Update run metadata
19. run-cancel - Cancel a running assistant execution
20. run-submit-tool-outputs - Submit tool call results to continue a run

Advanced Operations

21. run-step-list - List steps in a run execution
22. run-step-get - Get details of a specific run step

📚 MCP Resources Available

This server provides 9 comprehensive MCP resources to help you get started quickly:

🎯 Assistant Templates (4 resources)

• assistant://templates/coding-assistant - Pre-configured coding assistant
• assistant://templates/writing-assistant - Professional writing assistant
• assistant://templates/data-analyst - Data analysis assistant
• assistant://templates/customer-support - Customer support assistant

🔄 Workflow Examples (2 resources)

• examples://workflows/create-and-run - Complete workflow examples
• examples://workflows/batch-processing - Efficient batch processing

📖 Documentation (3 resources)

• docs://jezweb-mcp-core-api - Comprehensive API reference
• docs://error-handling - Common errors and solutions
• docs://best-practices - Guidelines for optimal usage

📖 Enhanced Usage Examples

Multi-Provider Usage

# Create an assistant (automatically selects best available provider)
"Create an assistant named 'Code Helper' with instructions to help with programming tasks"

# Use specific provider
"Create an assistant using OpenAI's GPT-4 model"
"Create an assistant using Anthropic's Claude model"

Assistant Management

# List all assistants
"List my assistants"

# Get assistant details
"Get details of assistant asst_abc123"

# Update an assistant
"Update assistant asst_abc123 to include the code_interpreter tool"

Thread and Message Management

# Create a new thread
"Create a new conversation thread"

# Add a message to a thread
"Add the message 'Hello, how can you help me?' to thread thread_abc123"

# List messages in a thread
"List all messages in thread thread_abc123"

Run Management

# Start an assistant run
"Start a run with assistant asst_abc123 on thread thread_abc123"

# Get run status
"Get status of run run_abc123"

# Cancel a running execution
"Cancel run run_abc123"

🔄 Deployment Option Parity

All deployment options provide identical functionality with all 22 tools working seamlessly:

✅ Functional Parity

• Identical Tools: All 22 tools work exactly the same way
• Same API Surface: Identical tool names, parameters, and responses
• Consistent Behavior: Error handling, validation, and responses are uniform
• Multi-Provider Support: All deployment options support multiple LLM providers

🚀 Transport Differences

🏗️ Architecture - Provider-Agnostic Design

Core Design Principles

• Adaptable - Support for multiple LLM providers through unified interface
• Simple - Environment-first configuration with sensible defaults
• Lightweight - Minimal dependencies and fast execution
• Extensible - Easy to add new providers and capabilities
• Reliable - Comprehensive error handling and fallback mechanisms

Provider System Architecture

// Provider Registry manages multiple LLM providers
interface LLMProvider {
  createAssistant(request: GenericCreateAssistantRequest): Promise<GenericAssistant>;
  listAssistants(request?: GenericListRequest): Promise<GenericListResponse<GenericAssistant>>;
  // ... all assistant API methods
}

// Providers implement the same interface
class OpenAIProvider implements LLMProvider { /* ... */ }
class AnthropicProvider implements LLMProvider { /* ... */ }
class GoogleProvider implements LLMProvider { /* ... */ }

Configuration System

Simple, environment-first configuration using standard environment variables:

# Required - at least one provider API key
export OPENAI_API_KEY="your-openai-key-here"
export ANTHROPIC_API_KEY="your-anthropic-key-here"

# Optional configuration
export JEZWEB_LOG_LEVEL="info"
export JEZWEB_DEFAULT_PROVIDER="openai"

The system automatically detects the deployment environment and applies appropriate defaults.

🧪 Testing Infrastructure

Comprehensive Test Suites

Both deployment options include robust testing:

NPM Package Testing

cd npm-package
npm test

Cloudflare Workers Testing

node test-validation-only.js

Manual Testing

Test the Cloudflare Workers deployment:

# List available tools
curl -X POST "https://openai-assistants-mcp.jezweb.ai/mcp/YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Test the NPM Package:

echo '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}' | npx jezweb-mcp-core@latest

🔧 Development

Local Development

1. Clone and install:

git clone https://github.com/jezweb/openai-assistants-mcp.git
cd openai-assistants-mcp
npm install

2. Set up environment:

wrangler secret put OPENAI_API_KEY
wrangler secret put ANTHROPIC_API_KEY

3. Start development:

npm run dev

Adding New Providers

1. Implement the LLMProvider interface
2. Create a provider factory
3. Register with the provider registry
4. Add configuration schema

Example:

class MyCustomProvider implements LLMProvider {
  // Implement all required methods
}

const factory: LLMProviderFactory = {
  create: (config) => new MyCustomProvider(config),
  getMetadata: () => ({ name: 'my-provider', ... }),
  validateConfig: (config) => true
};

registry.registerFactory(factory);

🔍 Enhanced Validation & Error Handling

Intelligent Error Messages

• Format Examples: Error messages include correct format examples
• Documentation References: Errors link to relevant documentation
• Suggestion Guidance: Invalid values show supported alternatives
• Provider Context: Errors include provider-specific guidance

Validation Features

• ID Format Validation: Strict format checking with helpful messages
• Provider Validation: Validates provider availability and capabilities
• Configuration Validation: Comprehensive config validation
• Parameter Validation: Type and range checking with examples

🔒 Security

• API Key Protection - Secure handling of multiple provider API keys
• Enhanced Input Validation - Comprehensive validation with helpful feedback
• Provider Isolation - Each provider operates in isolation
• CORS Security - Proper CORS headers for web clients
• Rate Limiting - Inherits provider-specific rate limits

🚀 Performance

• Global Edge - Deployed on Cloudflare's global network
• Sub-100ms - Typical response times under 100ms
• Provider Selection - Smart provider selection for optimal performance
• Efficient - Minimal memory footprint and fast execution

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

MIT License - see LICENSE for details.

🎯 Migration Guide

From OpenAI Assistants MCP v2.x

The migration is seamless - just update your package name:

# Old
npx openai-assistants-mcp@latest

# New
npx jezweb-mcp-core@latest

All existing tools and functionality remain identical. The new version adds multi-provider support while maintaining 100% backward compatibility.

Configuration Migration

Old environment variables continue to work:

# Still supported
OPENAI_API_KEY=your-key-here

# New multi-provider support
OPENAI_API_KEY=your-openai-key
ANTHROPIC_API_KEY=your-anthropic-key

Ready to get started? Choose your preferred installation method from the Quick Start guide above and begin building with multiple LLM providers through a unified interface!