Unleash the Power of Contextual AI with UBOS and Remote MCP Servers on Cloudflare

In the rapidly evolving landscape of Artificial Intelligence, the ability to provide Large Language Models (LLMs) with relevant, real-time context is paramount. This is where the Model Context Protocol (MCP) comes into play. MCP standardizes how applications provide context to LLMs, enabling them to access and interact with external data sources and tools. UBOS is a full-stack AI Agent development platform, designed to bring the power of AI Agents to every business department. Our platform helps you orchestrate AI Agents, connect them with your enterprise data, build custom AI Agents with your LLM model, and even create sophisticated Multi-Agent Systems.

This document delves into the deployment and customization of a remote MCP server on Cloudflare, specifically focusing on an authentication-free setup. We’ll explore how this architecture facilitates seamless integration with platforms like the Cloudflare AI Playground and Claude Desktop, empowering you to leverage AI tools effectively.

Understanding the Need for MCP Servers

LLMs, while powerful, are inherently limited by the data they were trained on. They lack the ability to access and process real-time information or interact with external systems. An MCP server acts as a bridge, allowing AI models to access and interact with external data sources and tools. This bridge is crucial for building AI agents that can perform complex tasks, make informed decisions, and adapt to changing environments. Without a robust context delivery mechanism like MCP, LLMs remain isolated and unable to fully realize their potential.

Deploying a Remote, Authentication-Free MCP Server on Cloudflare

This section details the process of deploying a remote MCP server on Cloudflare Workers without requiring authentication. This approach offers several advantages, including simplified setup, reduced operational overhead, and seamless scalability.

Getting Started

Cloudflare provides a one-click deployment option via the “Deploy to Workers” button, streamlining the setup process. Clicking this button redirects you to a deployment workflow that automatically configures and deploys the MCP server to your Cloudflare account.

Alternatively, you can use the Cloudflare command-line interface (CLI) to create the MCP server locally. This approach offers greater flexibility and control over the deployment process.

bash npm create cloudflare@latest – my-mcp-server --template=cloudflare/ai/demos/remote-mcp-authless

This command scaffolds a new Cloudflare Workers project based on the remote-mcp-authless template, providing a pre-configured MCP server that you can customize to your specific needs.

Customizing Your MCP Server

The core of your MCP server lies in the src/index.ts file. This file defines the tools that your AI models can access and interact with. To add your own tools, you’ll need to modify the init() method within this file.

Within the init() method, you can use the this.server.tool(...) function to register new tools with the MCP server. Each tool definition should include the tool’s name, a description of its functionality, and the code that executes the tool’s logic.

For example, to add a simple calculator tool, you might define it as follows:

typescript this.server.tool({ name: “calculator”, description: “Performs basic arithmetic operations.”, fn: async (query: string) => { try { // Parse the query and perform the calculation const result = eval(query); return result.toString(); } catch (error) { return “Error: Invalid calculation.”; } }, });

This example demonstrates how to define a tool that accepts a mathematical expression as input, evaluates it using the eval() function, and returns the result.

Connecting to Cloudflare AI Playground

The Cloudflare AI Playground provides a convenient interface for interacting with your MCP server. To connect to your server from the Playground, follow these steps:

1. Navigate to the Cloudflare AI Playground.
2. Enter the URL of your deployed MCP server (e.g., remote-mcp-server-authless.<your-account>.workers.dev/sse).
3. The Playground will automatically connect to your MCP server and display the available tools.

From the Playground, you can now invoke your MCP tools directly, providing input and receiving output in real-time. This allows you to experiment with different AI models and tools, and to quickly prototype new AI-powered applications.

Connecting Claude Desktop to Your MCP Server

Claude Desktop, a popular AI assistant, can also be configured to connect to your remote MCP server. This allows you to leverage your custom tools within the Claude Desktop environment.

To connect Claude Desktop to your MCP server, follow these steps:

1. Follow Anthropic’s Quickstart guide to set up Claude Desktop.
2. Open Claude Desktop and navigate to Settings > Developer > Edit Config.
3. Update the configuration file with the following JSON:

{ “mcpServers”: { “calculator”: { “command”: “npx”, “args”: [ “mcp-remote”, “http://localhost:8787/sse” // or remote-mcp-server-authless.your-account>.workers.dev/sse ] } } }

Replace http://localhost:8787/sse with the actual URL of your deployed MCP server. Restart Claude Desktop, and the tools defined on your MCP server should now be available within the Claude environment.

Use Cases for Remote MCP Servers

• Contextual AI Agents: Build AI agents that can access and process real-time information from various sources, enabling them to make more informed decisions and perform complex tasks.
• AI-Powered Automation: Automate tasks by integrating AI models with external systems and tools, allowing them to perform actions based on real-time data and user input.
• Enhanced Customer Service: Provide customers with personalized and relevant support by integrating AI models with CRM systems and knowledge bases.
• Data-Driven Insights: Extract valuable insights from data by integrating AI models with data analytics tools and databases.

Key Features and Benefits

• Simplified Deployment: One-click deployment to Cloudflare Workers simplifies the setup process.
• Scalability: Cloudflare Workers provides automatic scalability, ensuring that your MCP server can handle increasing traffic.
• Cost-Effectiveness: Cloudflare Workers offers a pay-as-you-go pricing model, making it a cost-effective solution for deploying MCP servers.
• Customizability: The src/index.ts file allows you to customize the MCP server by adding your own tools and integrations.
• Integration with Popular Platforms: Seamless integration with Cloudflare AI Playground and Claude Desktop enables you to leverage your MCP tools within familiar environments.

UBOS: Empowering AI Agent Development

UBOS takes the power of MCP servers and elevates it to a comprehensive AI Agent development platform. With UBOS, you can:

• Orchestrate AI Agents: Design and manage complex workflows involving multiple AI agents, enabling them to collaborate and solve complex problems.
• Connect to Enterprise Data: Seamlessly integrate AI agents with your enterprise data sources, providing them with the context they need to make informed decisions.
• Build Custom AI Agents: Develop custom AI agents tailored to your specific needs, using your own LLM models and data.
• Create Multi-Agent Systems: Build sophisticated multi-agent systems that can autonomously perform complex tasks and adapt to changing environments.

By combining the power of MCP servers with the comprehensive capabilities of UBOS, you can unlock the full potential of AI and transform your business.

In conclusion, deploying a remote, authentication-free MCP server on Cloudflare offers a streamlined and cost-effective way to integrate AI models with external data sources and tools. By leveraging the flexibility of Cloudflare Workers and the integration capabilities of platforms like the Cloudflare AI Playground and Claude Desktop, you can build powerful AI-powered applications that drive innovation and efficiency. And with UBOS, you can take your AI agent development to the next level, creating sophisticated and intelligent systems that can solve real-world problems.