Documentation

The Model Context Protocol (MCP) is an open protocol designed to standardize how AI applications, particularly those using Large Language Models (LLMs), access external data and tools. Think of MCP as a universal adapter for AI: it provides a consistent way for LLMs to connect to diverse information sources and functional capabilities.

Why is MCP Important?

• Enhanced Agent Capabilities: Enables the development of more sophisticated AI agents and complex workflows by integrating LLMs with real-time data and external tools.
• Interoperability: Offers flexibility to switch between different LLM providers and vendors without overhauling integrations.
• Standardized Security: Promotes best practices for securing data access within your existing infrastructure.

What is mcpDploy?

mcpDploy is a Software-as-a-Service (SaaS) platform that empowers you to create, configure, and deploy custom MCP servers with **zero coding required**. We abstract away the complexities of server setup, infrastructure management, and deployment. Simply define your server's behavior through our intuitive web interface, and mcpDploy provides you with an instant, production-ready MCP endpoint and API key.

1. Sign Up / Log In

Begin by creating an account or logging into your existing mcpDploy account. This grants you access to your personal dashboard where you'll manage all your MCP projects.

2. Dashboard Overview

Your mcpDploy dashboard is your central hub. It provides an at-a-glance view of your active servers, API call statistics (feature coming soon), and user engagement metrics (feature coming soon). From the dashboard, you can initiate the creation of new MCP servers or access and manage your existing deployments.

Building an MCP server on mcpDploy is a guided, multi-step process designed for clarity and ease:

1. Basic Information: Start by giving your server a unique name and version (e.g., my-chat-agent-v1.0.0). You can also add an optional description for better organization.
2. Resources (Optional): Define data resources that your MCP server can expose. Each resource requires a name, a URI pattern (e.g., /api/user_profile/{userId}), the backend API URL that will serve the data, and any necessary HTTP headers for authentication or content type.
3. Tools (Optional): Configure tools (callable functions) that your MCP server makes available. This involves specifying a tool name, a clear description of its purpose, the API URL for the tool's backend logic, the appropriate HTTP method (GET, POST, etc.), and any required headers or parameter definitions that the tool expects.
4. Prompts (Optional): Set up reusable prompt templates to standardize interactions. Each prompt includes a name, description, the template string (which can use placeholders like {{user_query}}), and definitions for any parameters used within the template.
5. Review & Deploy: Carefully review your entire server configuration. Once you're satisfied, click "Deploy Server". mcpDploy will instantly provision your MCP server and provide you with a unique endpoint URL and API key.

The session_management flag controls whether your MCP server maintains state across multiple requests. When enabled, mcpDploy preserves context—such as previous tool outputs and user prompts—enabling stateful, multi-turn interactions. When disabled, each request is handled independently (stateless).

Use session management to build complex, conversational workflows that rely on context persistence. If you prefer a simpler, stateless setup for one-off calls or for performance reasons, disable this flag.

Once your MCP server is live, you can manage it effectively from your mcpDploy dashboard:

• Access detailed information for each server, including its name, version, description, unique endpoint URL, and associated API key.
• Review the specific configurations for all associated resources, tools, and prompts.
• Conveniently copy the endpoint URL and API key for easy client-side configuration.
• Permanently delete servers when they are no longer needed (note: this action cannot be undone).

After deploying your server on mcpDploy, you'll receive an endpoint URL and an API key. Use these to configure your MCP clients. Below are examples for common clients. Replace placeholder URLs and API keys with your actual server details found in your mcpDploy dashboard.

{
  "mcpServers": {
    "My-mcpDploy-Server": {
      "transport": "sse",
      "url": "https://your-mcpdploy-endpoint.com/mcp/your-server-id",
      "headers": {
        "Accept": "text/event-stream",
        "Content-Type": "application/json",
        ""X-API-Key"": "Bearer YOUR_API_KEY_FROM_MCPDPLOY"
      }
    }
  }
}

{
  "modelContextProtocol.servers": {
    "my_mcpdploy_server": {
      "url": "https://your-mcpdploy-endpoint.com/mcp/your-server-id",
      "headers": {
        "Accept": "application/json, text/event-stream",
        "Content-Type": "application/json",
        ""X-API-Key"": "Bearer YOUR_API_KEY_FROM_MCPDPLOY"
      }
    }
  }
}

{
  "mcpServers": {
    "my_mcpdploy_server": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://your-mcpdploy-endpoint.com/mcp/your-server-id",
        "--header", "Authorization: Bearer YOUR_API_KEY_FROM_MCPDPLOY"
      ]
    }
  }
}

{
  "mcpServers": {
    "my_mcpdploy_tool_server": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://your-mcpdploy-endpoint.com/mcp/your-server-id",
        "--header", "Authorization: Bearer YOUR_API_KEY_FROM_MCPDPLOY"
      ]
    }
  }
}

• Resources: These represent specific data sources your MCP server can access. They effectively map a URI pattern on your MCP server (e.g., /news/latest) to an external API that provides the actual data. You configure the backend URL and any necessary authentication headers for mcpDploy to fetch this data on behalf of the LLM.
• Tools: These represent functions or actions that your LLM can invoke through the MCP server. Each tool maps a unique name (e.g., send_confirmation_email) to an external API endpoint that performs the action. You define the HTTP method, parameters, and headers required by the backend service.
• Prompts: These are reusable text templates that can be dynamically populated with parameters. Prompts help in standardizing interactions with LLMs and can be fetched by clients to guide conversations or tasks. For example, a prompt for summarization might be Summarize the following text: {{text_to_summarize}}.