Toradex Developer Center Model Context Protocol (MCP) Server

The Toradex Developer Center Model Context Protocol (MCP) server gives compatible AI clients access to Toradex technical information. You can use it to search Toradex documentation and public sources or ask the Toradex AI Support Agent a technical question.

The server provides the following remote MCP endpoint over Streamable HTTP:

https://mcp.toradex.com/mcp

The endpoint does not require an API key.

Available Tools

The server provides the following tools to your AI client:

Tool	Purpose
search-toradex-docs	Performs a semantic search of Toradex documentation and other public sources, such as Toradex repositories on GitHub. It returns relevant extracts and source URLs.
ask-question-about-toradex	Sends a troubleshooting, capability, or conceptual question to the Toradex AI Support Agent and returns its answer.

Your AI client selects a tool based on your prompt. You can also name a tool in your prompt if you want the client to use a specific one.

Prerequisites

To use the server, you need:

• An MCP-compatible client that supports remote servers over Streamable HTTP.
• Internet access to https://mcp.toradex.com over TCP port 443.

The configuration steps depend on your MCP client. Use the endpoint URL in your client's remote MCP server settings. This article provides examples for Visual Studio Code, Cursor, Windsurf, OpenAI clients, Claude.ai, Claude Code, and Claude Desktop.

Use the Developer Center Setup Shortcuts

On a desktop browser, you can copy or install the MCP configuration from any Toradex Developer Center article:

1. Select Copy page near the table of contents.

2. Select one of the following options:

   • Copy MCP config copies the Toradex MCP server JSON. Paste the server entry into the MCP configuration for your client.
   • Install in Cursor opens the native MCP installer in Cursor.
   • Install in VS Code opens the native MCP installer in Visual Studio Code.

3. Review and approve the server configuration in your client.

4. Confirm that the Toradex tools appear in the client's MCP server or tools list.

If your browser does not open an editor from an installation link, use the manual configuration for that client in this article.

Configure Visual Studio Code

Visual Studio Code stores workspace MCP server configuration in .vscode/mcp.json.

1. Create .vscode/mcp.json in your workspace.

2. Add the Toradex server:

   .vscode/mcp.json

   {
     "servers": {
       "toradex": {
         "type": "http",
         "url": "https://mcp.toradex.com/mcp"
       }
     }
   }

3. Start the toradex server from the MCP server list in Visual Studio Code.

4. Open the tools picker in chat and confirm that the Toradex tools are available.

For details about workspace and user configuration, refer to the Visual Studio Code MCP server documentation.

Configure Cursor

1. Create the .cursor directory in your project root if it does not exist.

2. Create .cursor/mcp.json with the following configuration:

   .cursor/mcp.json

   {
     "mcpServers": {
       "toradex": {
         "url": "https://mcp.toradex.com/mcp"
       }
     }
   }

3. Save the file and restart Cursor if the server does not appear immediately.

To make the server available in every project, open the command palette and select Cursor Settings > MCP > Add new global MCP server instead.

Configure Windsurf

Add the server to your Windsurf mcp.json file. Windsurf uses the serverUrl property for a remote MCP server:

mcp.json

{
  "mcpServers": {
    "toradex": {
      "serverUrl": "https://mcp.toradex.com/mcp"
    }
  }
}

Save the file and refresh the MCP server list in Windsurf.

Configure OpenAI Clients

OpenAI provides different MCP setup options for Codex and ChatGPT:

• Codex and the ChatGPT desktop app can connect directly to an MCP server.
• ChatGPT individual accounts can use the ChatGPT desktop app or the plugin options available to their account.
• ChatGPT Enterprise workspaces can control access to MCP-backed plugins and tools through workspace policy.

For complete setup steps and availability, refer to the OpenAI MCP documentation for Codex and the ChatGPT desktop app and the OpenAI plugin connection documentation for ChatGPT web.

Use the following values when you add the Toradex server:

• Name: Toradex Developer Center MCP
• Transport: Streamable HTTP
• URL: https://mcp.toradex.com/mcp

Refer to the linked OpenAI documentation for the exact configuration file or CLI syntax for your client.

Configure Claude.ai

Claude.ai supports remote MCP servers as custom connectors.

• Individual accounts add a connector from Settings > Customize > Connectors.
• Team and Enterprise accounts require an **organization Owner ** to register the connector before members can connect.

For complete setup steps and network requirements, refer to the Anthropic custom connector documentation.

Use the following values when you add the Toradex connector:

• Name: Toradex Developer Center MCP
• URL: https://mcp.toradex.com/mcp

Leave the OAuth client ID, client secret, and request headers empty. The Toradex endpoint does not require authentication.

For example, ask Claude:

Using the Toradex connector, which BSP versions support Verdin iMX8M Plus? Include the source links.

Search the Toradex developer resources for instructions on enabling CAN on Verdin AM62.

Configure Claude Code

Claude Code calls the Streamable HTTP transport http. This name describes the MCP transport and does not mean that the connection is unencrypted. The endpoint URL uses HTTPS and the connection is encrypted with TLS.

Add the Server from the Command Line

Add the remote server from a host computer terminal:

$ claude mcp add --transport http toradex https://mcp.toradex.com/mcp

Verify the configuration:

$ claude mcp list

You should see an output similar to the following:

toradex: https://mcp.toradex.com/mcp (HTTP) - ✔ Connected

Add the Server from JSON

If you prefer the MCP specification's streamable-http name, add the same server using JSON:

$ claude mcp add-json toradex '{"type":"streamable-http","url":"https://mcp.toradex.com/mcp"}'

You can also share the server configuration with a project by adding a .mcp.json file at the project root:

.mcp.json

{
  "mcpServers": {
    "toradex": {
      "type": "streamable-http",
      "url": "https://mcp.toradex.com/mcp"
    }
  }
}

Claude Code asks each user to approve project-scoped MCP servers before using them. Use a user-scoped command instead if you want the server available across all projects without committing .mcp.json:

$ claude mcp add --transport http toradex --scope user https://mcp.toradex.com/mcp

Claude Code can then select the Toradex tools when a prompt requires Toradex technical information. For more configuration options, refer to the Claude Code MCP documentation.

Configure Claude Desktop

Use the mcp-remote adapter to connect Claude Desktop to the remote endpoint. This option requires Node.js and downloads the adapter through npx when Claude Desktop starts it.

Add the following server to claude_desktop_config.json:

claude_desktop_config.json

{
  "mcpServers": {
    "toradex": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.toradex.com/mcp"
      ]
    }
  }
}

Restart Claude Desktop after saving the file. Use a client's native Streamable HTTP configuration when it is available; the local adapter is only necessary for clients that cannot connect to a remote HTTP MCP server directly.

Verify the Connection

Open your client's MCP server or tools list and confirm that the Toradex server is connected. Then ask:

Search the Toradex documentation for the production-hardening recommendations for Torizon OS and return the source links.

The client should show that it called search-toradex-docs. A successful connection also exposes ask-question-about-toradex.

Encourage the Client to Use Toradex Tools

MCP clients normally select tools automatically. If your client supports user rules and does not reliably select the Toradex tools, add an instruction similar to the following:

When answering questions about Toradex products or documentation, use the Toradex MCP tools. Use search-toradex-docs to find documentation, GitHub issues, release notes, and community content. Use ask-question-about-toradex for specific troubleshooting, product, API, or implementation questions.

Use the Server

Ask a focused question and include the product, software version, and relevant error messages when applicable. For example:

Search the Toradex documentation for the production-hardening recommendations for Torizon OS. Summarize the checklist and include the source links.

Ask the Toradex support agent how to troubleshoot a Verdin iMX8M Plus that does not boot after a device tree change. Include the diagnostic commands I should run.

Find the documentation for enabling the Cortex-M core on an Apalis iMX8 and explain the prerequisites.

Review the returned source links before applying commands or configuration changes to a device. Search results contain extracts, and an AI-generated answer might not include all product-specific constraints.

warning

Prompts and the context your client sends to a remote tool leave your local environment. Do not include credentials, private keys, customer data, or other confidential information in a request.

Troubleshooting

The Endpoint Returns HTTP 400 in a Browser

This behavior is expected. An empty-looking response is also possible. The MCP endpoint accepts initialization and protocol messages from an MCP client, not regular browser navigation. Configure the URL in an MCP client and verify the connection from that client.

The Client Cannot Connect

Check the following items:

• Use the complete URL https://mcp.toradex.com/mcp. Do not omit https://.
• In Claude Code, --transport http selects Streamable HTTP; it does not change the HTTPS endpoint to an unencrypted connection.
• Select Streamable HTTP or HTTP as the transport. Do not configure the endpoint as a local stdio server or use a legacy /sse URL.
• Confirm that your network, proxy, or firewall permits HTTPS connections to the configured endpoint over TCP port 443.
• For Claude.ai, remember that Anthropic's infrastructure connects to the endpoint. A successful connection from your computer does not confirm that a firewall or CDN permits the Claude.ai connection.
• Confirm that your MCP client supports remote Streamable HTTP servers.

Streamable HTTP is the standard transport for a remote MCP server. Legacy HTTP with SSE is deprecated. The stdio transport requires a locally installed process, so it is not a direct alternative for connecting to the hosted Toradex endpoint.

Do not use ping as an availability test. Cloud load balancers commonly block ICMP even while HTTPS is working. Verify the connection from the MCP client or test TCP port 443 instead.

Claude Code Offers Authentication

The public Toradex MCP endpoint does not require OAuth. If Claude Code displays Dynamic Client Registration rejected (HTTP 404), do not select Authenticate. Confirm the HTTPS URL, clear any stale server entry, add it again, and select Reconnect from the /mcp menu.

The Toradex Tools Do Not Appear

Restart or reconnect the server from your MCP client after changing its configuration. Then inspect the client logs for connection or protocol errors. A successful connection exposes search-toradex-docs and ask-question-about-toradex in the client's tool list.

The Answer Is Too General

Add the exact Toradex module, carrier board, operating system, release version, and observed behavior to your prompt. Ask the client to return source links so that you can review the relevant documentation in context.