Unleashing the Power of Context: Remote MCP Server on Cloudflare with UBOS

In the rapidly evolving landscape of AI, context is king. Large Language Models (LLMs) like GPT-4 and Claude are powerful, but their true potential is unlocked when they can access and leverage external data sources and tools. This is where the Model Context Protocol (MCP) comes in, providing a standardized way for applications to provide this crucial context to LLMs.

This document outlines how to set up a remote MCP server on Cloudflare Workers, complete with OAuth login, enabling seamless integration with tools like the MCP Inspector and, crucially, Claude. We’ll delve into the development process, deployment, and debugging, all while highlighting how this approach synergizes with the UBOS platform for streamlined AI agent development.

Understanding the MCP Paradigm

The Model Context Protocol (MCP) standardizes how applications provide context to LLMs. Think of it as a universal translator, allowing AI models to understand and interact with the real world through structured data and predefined actions. An MCP server acts as the central hub, managing connections and orchestrating the flow of information between the LLM and external resources.

By abstracting away the complexities of data retrieval and tool invocation, MCP empowers developers to focus on building intelligent AI agents that can perform a wide range of tasks, from answering complex questions to automating business processes.

Use Cases: Beyond Simple Chatbots

The ability to provide rich context opens up a plethora of use cases for LLMs. Here are just a few examples:

• Enhanced Customer Service: An AI agent connected to a CRM database via MCP can access customer information, order history, and support tickets to provide personalized and informed assistance.
• Data-Driven Decision Making: An AI agent integrated with business intelligence tools through MCP can analyze real-time data, identify trends, and provide actionable insights to decision-makers.
• Automated Content Creation: An AI agent connected to a content management system via MCP can generate articles, blog posts, and marketing materials based on specific keywords, topics, and target audiences.
• Proactive Cybersecurity: An AI agent leveraging threat intelligence feeds and security tools through MCP can detect and respond to potential security threats in real-time.
• Financial Analysis & Trading: Accessing real-time market data and analysis tools via MCP, an AI agent can assist in making informed investment decisions.

Key Features: Building Blocks for Intelligent Agents

The MCP server setup outlined in this document provides a foundation for building powerful AI agents with the following key features:

• Remote Accessibility: Deploying the MCP server on Cloudflare Workers ensures global availability and scalability.
• OAuth Authentication: Secure access to the MCP server is enforced through OAuth login, protecting sensitive data and preventing unauthorized access.
• Seamless Integration: The MCP server integrates seamlessly with tools like the MCP Inspector and Claude, simplifying development and testing.
• Standardized Protocol: Adherence to the MCP standard ensures interoperability with a wide range of LLMs and external resources.
• Customizable Tools: Developers can define custom tools and actions that the AI agent can invoke, tailoring its capabilities to specific use cases.

Setting Up Your Remote MCP Server: A Step-by-Step Guide

The following steps outline the process of setting up a remote MCP server on Cloudflare Workers:

1. Local Development:

   • Clone the repository: git clone git@github.com:cloudflare/ai.git
   • Install dependencies: cd ai && npm install
   • Run locally: npx nx dev remote-mcp-server
   • Access the server at http://localhost:8787/

2. Connecting the MCP Inspector:

   • Start the MCP Inspector: npx @modelcontextprotocol/inspector
   • Configure the Inspector to use SSE transport and connect to http://localhost:8787/sse
   • Log in with any email and password.
   • Explore the available tools.

3. Connecting Claude Desktop:

   • Follow Anthropic’s Quickstart guide.

   • Edit the Claude configuration file (Settings > Developer > Edit Config).

   • Replace the configuration with the following:

     { “mcpServers”: { “math”: { “command”: “npx”, “args”: [ “mcp-remote”, “http://localhost:8787/sse” ] } } }

4. Deployment to Cloudflare:

   • Create a KV namespace: npx wrangler kv namespace create OAUTH_KV
   • Add the KV namespace ID to wrangler.jsonc.
   • Deploy: npm run deploy

5. Remote Access:

   • Use the MCP Inspector to connect to the deployed server using the workers.dev URL (e.g., worker-name.account-name.workers.dev/sse).
   • Update the Claude configuration file to point to the deployed server’s URL and restart Claude.

Debugging Tips and Tricks

• If you encounter issues, try restarting Claude.
• Connect directly to the MCP server using the command line: npx mcp-remote http://localhost:8787/sse
• In rare cases, clearing the files in ~/.mcp-auth may help: rm -rf ~/.mcp-auth

Integrating with the UBOS Platform: Streamlining AI Agent Development

The process described above can be significantly streamlined and enhanced by leveraging the UBOS platform. UBOS is a full-stack AI Agent Development Platform that empowers businesses to orchestrate AI Agents, connect them with enterprise data, build custom AI Agents with their own LLM models, and create sophisticated Multi-Agent Systems.

Here’s how UBOS can enhance your MCP server deployment:

• Simplified Deployment: UBOS provides a user-friendly interface for deploying and managing AI agents and their dependencies, including MCP servers, reducing the complexity of manual configuration and deployment.
• Centralized Management: UBOS offers a centralized platform for managing all your AI agents, data sources, and tools, providing a single pane of glass for monitoring performance, troubleshooting issues, and scaling your AI infrastructure.
• Enhanced Security: UBOS provides robust security features, including access control, data encryption, and audit logging, ensuring the security and compliance of your AI applications.
• Scalability and Reliability: UBOS is built on a scalable and reliable infrastructure, ensuring that your AI agents can handle increasing workloads and maintain high availability.
• Data Integration: UBOS simplifies the process of connecting your AI agents to various data sources, including databases, APIs, and cloud storage, enabling them to access the information they need to perform their tasks effectively. UBOS provides tools that help you connecting your enterprise data.
• AI Agent Orchestration: UBOS provides advanced orchestration capabilities, allowing you to define complex workflows and interactions between multiple AI agents, creating sophisticated Multi-Agent Systems that can solve complex problems.
• Custom AI Agent Development: Build custom AI Agents with your LLM model.

By integrating your MCP server with the UBOS platform, you can unlock the full potential of AI and accelerate your AI initiatives.

Conclusion: Embracing the Future of AI with Context and UBOS

Setting up a remote MCP server on Cloudflare Workers is a crucial step towards building intelligent and context-aware AI agents. By following the steps outlined in this document and leveraging the power of the UBOS platform, you can empower your AI agents to access and leverage external data sources and tools, unlocking a wide range of use cases and driving innovation across your organization. Embrace the future of AI with context and UBOS.