Rocket MCP

The Model Context Protocol (MCP) is an open standard that lets AI tools connect to external services and take actions via structured tool calls. Rocket MCP connects your Rocket.net account to MCP-capable clients so you can manage Rocket resources through conversation instead of clicking through dashboards.

Why use MCP

Rocket MCP allows access to do anything you can do through the Rocket.net API but without having to write a single line of code. Good news, the Rocket.net platform is 100% API-driven so Rocket MCP has the power to do anything you need. You can simply ask Rocket MCP questions or give it commands and you can do many amazing things with your sites.

NOTE: With great power comes great responsibility. Make Peter Parker proud and be careful to make changes against staging rather than production to make sure you don't affect your live site in ways you don't intend.

Examples:

Requirements

Need an account? Sign up at rocket.net/registration.php.

Get an access token (JWT)

Some clients (like Gemini CLI) are easiest to configure with a bearer token. You can generate a JWT by following the steps below.

How Rocket MCP works (high level)

Rocket exposes an MCP server over HTTP + Server-Sent Events (SSE). Many AI clients only support MCP servers that run as a local process, so we use mcp-remote as a local bridge that connects to Rocket’s SSE URL.

Step 1: Choose your authentication method

Rocket MCP supports passing credentials as request headers. The recommended approach is to provide ROCKETNET_USERNAME and ROCKETNET_PASSWORD headers so the server can obtain/refresh JWTs as needed.

Treat credentials like passwords: don’t commit them, don’t paste them into logs, and rotate them if they leak.

Step 2: Connect your AI client (recommended)

Add the following MCP server configuration to your client. This runs a local mcp-remote process which connects to Rocket’s SSE endpoint: https://mcp-staging.onrocket.cloud/sse.

{
  "mcpServers": {
    "rocket-net-sse": {
      "command": "npx",
      "args": [
        "mcp-remote", 
        "https://mcp-staging.onrocket.cloud/mcp", 
        "--transport",
        "http",
        "--header", "Authorization: Bearer YOUR_ACCESS_TOKEN_GOES_HERE"
      ]
    }
  }
}

Client-specific setup

Different clients store MCP settings in different places, but the goal is always the same: run mcp-remote locally and point it at Rocket’s SSE URL, passing your credentials as headers.

Cursor

In Cursor, add Rocket MCP through the MCP settings UI:

  1. Open the command palette:
    • macOS: Cmd + Shift + P
    • Windows: Ctrl + Shift + P
  2. Select View: Open MCP Settings
  3. Add a new custom MCP server by pasting the configuration from Step 2, then save the mcp.json file.

Claude Desktop

Claude Desktop reads a local JSON config file. Add a mcpServers entry that runs mcp-remote.

Common config locations:

After updating the file, fully quit and reopen Claude Desktop.

Claude Code (terminal)

Claude Code typically manages MCP servers via a CLI command. If your install supports it, add Rocket MCP using an HTTP transport and pass headers for auth.

# Example shape (adjust to your CLI's exact flags):
claude mcp add --transport http rocket-net https://mcp-staging.onrocket.cloud/mcp \
  --header "Authorization: Bearer YOUR_ACCESS_TOKEN_GOES_HERE"

If your Claude Code build only supports local processes (no remote URLs), use the mcp-remote JSON config pattern shown above instead.

Gemini CLI

Gemini CLI can be configured with a direct MCP server URL and an Authorization bearer token. First, generate a JWT (see Getting started), then add it to your Gemini MCP config.

{
  "security": {
    "auth": {
      "selectedType": "oauth-personal"
    }
  },
  "mcpServers": {
    "rocket-net": {
      "url": "https://mcp-staging.onrocket.cloud/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN_HERE"
      }
    }
  }
}

Other clients

If your client supports MCP but you don’t see Rocket-specific docs for it, use one of these two patterns:

Step 3: Confirm it’s working

After saving your MCP settings, restart your AI client (or reload its window), then try a simple request like:

Troubleshooting