MCP Service

RequestRocket exposes a native Model Context Protocol (MCP) service, allowing AI agents — such as Cursor, Claude Desktop, and Claude.ai — to manage your proxies, credentials, targets, rules, and more entirely through natural language.

Once connected, an AI agent can create and configure proxies, inspect request history, execute test requests, and troubleshoot your integration without you needing to leave your editor or chat window.

Prerequisites

Before connecting an AI agent you will need:

• A RequestRocket account
• A user API token (u-...) — found under Settings → Keys in the dashboard

MCP Endpoint

All requests must be sent as POST to the endpoint above. GET and other methods return 405.

Authentication

The MCP service accepts two authentication methods. For most users connecting Cursor or Claude Desktop, a user API token is the simplest option.

For OAuth, the server publishes standard discovery metadata:

• GET https://api.requestrocket.com/.well-known/oauth-protected-resource
• GET https://api.requestrocket.com/.well-known/oauth-authorization-server

OAuth-capable clients discover these endpoints automatically — no manual configuration is needed beyond pointing the client at the MCP URL.

Connecting Cursor

{
  "mcpServers": {
    "requestrocket": {
      "url": "https://api.requestrocket.com/mcp",
      "headers": {
        "Authorization": "u-your-api-token-here"
      }
    }
  }
}

Connecting Claude Desktop

~/Library/Application Support/Claude/claude_desktop_config.json

%APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "requestrocket": {
      "type": "http",
      "url": "https://api.requestrocket.com/mcp",
      "headers": {
        "Authorization": "u-your-api-token-here"
      }
    }
  }
}

Connecting Claude.ai (OAuth)

Claude.ai supports OAuth-based MCP connections which do not require you to manually manage a token.

https://api.requestrocket.com/mcp

Available Capabilities

The MCP server exposes the full RequestRocket platform surface through approximately 90 tools, grouped by domain:

Resources

In addition to tools, the server provides MCP resources the agent can read:

• requestrocket://schemas/credentials/<type> — JSON Schema for each credential type
• requestrocket://schemas/proxy/create — JSON Schema for proxy creation
• requestrocket://examples/responses/* — Example API response payloads
• requestrocket://client/{clientId}/targets — Live list of targets for a client
• requestrocket://client/{clientId}/proxies — Live list of proxies for a client
• requestrocket://client/{clientId}/credentials — Live list of credentials for a client

Proxy Setup via Agent

When creating a new proxy through an AI agent, the recommended setup order is automatically guided by the server instructions:

1. create_target        → Define the upstream API endpoint
2. create_credential    → Create a proxy credential (how callers authenticate to you)
3. create_credential    → Create a target credential (how RequestRocket authenticates to upstream)
4. create_proxy         → Combine the above into a proxy
5. execute_proxy_read_request → Send a test request to confirm everything works

Simply tell the agent what you want to connect to, and it will walk through these steps with you.

Safety & Best Practices

Additional safety rules enforced by the server:

• The agent will warn you before setting a proxy or rule to inactive, as this immediately stops live traffic
• Resource IDs are never fabricated — the agent always retrieves them from live list/get calls
• Billing, subscription management, member invites, and password resets are always directed to the web UI

Troubleshooting