Dwellsy MCP Server

Welcome to the Dwellsy MCP Server. This server leverages the modern Streamable HTTP (2025-03-26) protocol to provide high-throughput, low-latency tool and context execution for AI agents via the Model Context Protocol (MCP).

This guide explains how to connect your AI agent to this server using two methods: a direct remote connection and a local proxy connection.

Method 1: Direct Remote Connection (Recommended)

If your AI agent supports remote MCP connections (typically via Server-Sent Events - SSE), you can connect directly to the hosted server URL. This provides the best performance by utilizing native Streamable HTTP capabilities.

Configuration Example (JSON)

Add the following to your AI agent's MCP configuration file (e.g., mcp_settings.json):

{
  "mcpServers": {
    "dwellsy-mcp": {
      "type": "http",
      "url": "https://mcp.dwellsy.com/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_TOKEN"
      }
    }
  }
}
Note: Ensure that your network allows outbound persistent HTTP connections, as Streamable HTTP relies on long-lived streams for agent-server communication.

Method 2: Local Proxy Connection (stdio to Remote Bridge)

Many desktop AI agents (like Claude Desktop) currently only support local MCP servers via the stdio transport. To connect these agents to Dwellsy MCP server, you must use a local proxy wrapper.

The proxy runs locally on your machine, communicates with the AI agent via stdio, and forwards the MCP protocol messages over the internet to your remote server.

Configure the AI Agent

Configure your local agent to run the proxy command, passing Dwellsy MCP server URL as an argument.

Claude Desktop Configuration (claude_desktop_config.json)

{
  "mcpServers": {
    "dwellsy-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.dwellsy.com/api/mcp",
        "--header",
        "Authorization: Bearer ${AUTH_TOKEN}"
      ],
      "env": {
        "AUTH_TOKEN": "YOUR_API_TOKEN"
      }
    }
  }
}

How It Works

  1. The AI Agent spawns the npx proxy process.
  2. The Agent sends JSON-RPC messages via standard input (stdin).
  3. The local proxy catches these messages and streams them over HTTP to your remote server.
  4. The server responds via the HTTP stream, which the proxy prints back to standard output (stdout) for the Agent to read.

Method 3: Connecting via platform.openai.com

OpenAI's platform provides native support for remote MCP servers, allowing you to easily add your Streamable HTTP server to your custom GPTs, Agent Builders, and Playground environments.

Step-by-Step Integration

  1. Generate Auth Token: Open account page http://mcp-dev.dwellsy.com/tokens, enter a token name in "Base oAuth Client" section and click generate. Save it somewhere.
  2. Access the Platform: Navigate to platform.openai.com/chat (or the Playground/Agent Builder) and click + Create to start a new chat or agent.
  3. Add MCP Server: Locate the Tools section in the settings panel. Click + Add, and select MCP Server from the dropdown menu.
  4. Create New Connection: Click the + Server (or "Add New") button in the modal window.
  5. Configure Details: Fill out the configuration form:
    • URL: Enter your full Streamable HTTP server URL (e.g., https://mcp.dwellsy.com/api/mcp).
    • Authentication: Select Access token / API key from the dropdown.
    • Key: Paste your AUTH_TOKEN_FROM_STEP_1 here. You can skip this step for search listings tools.
    • Label: Give your server a recognizable name (e.g., "Streamable HTTP Tools").
  6. Connect and Discover: Click Connect. OpenAI will connect to your Streamable HTTP server and automatically discover the available tools.
  7. Finalize: Review the list of fetched tools and click Add to make them available to your agent. You can now prompt the LLM to use these tools natively!

Troubleshooting