SocialGO

Developers

MCP Server

Plug SocialGO into Claude, Cursor and any Model Context Protocol client so your AI agents can search the catalog and run orders on their own.

What is the SocialGO MCP server?

The SocialGO MCP server exposes the panel as a set of tools for AI agents over the Model Context Protocol. Once connected, an assistant can browse the service catalog, place and track orders, request refills and manage balance, with no custom glue code. It is a thin read-and-act layer on top of the same API your dashboard uses, so agents stay within the limits and pricing of your account.

It works with any MCP-capable client, including Claude Desktop, Claude Code, Cursor and your own agents built on an MCP SDK.

Add it to your client

The server ships as the @socialgo/mcp package and runs over stdio via npx, so there is nothing to install or self-host. Add the following entry to your MCP client configuration:

{
  "mcpServers": {
    "socialgo": {
      "command": "npx",
      "args": ["-y", "@socialgo/mcp"],
      "env": {
        "SOCIALGO_API_KEY": "your_api_key_here"
      }
    }
  }
}

With Claude Code you can register it in a single command from your terminal:

claude mcp add socialgo -- npx @socialgo/mcp

After adding it, remember to set SOCIALGO_API_KEY in the environment so the command above can pick it up.

Authentication

The server reads a single environment variable, SOCIALGO_API_KEY. Generate a key from your account settings, keep it secret and never commit it to source control. Every tool call is authenticated and authorized with this key, and orders are billed against the matching account balance. Rotate the key from your dashboard at any time to revoke access.

Core tools

The server keeps a small, stable surface of tools that cover the full order lifecycle:

  • search_servicesFind services by platform, type or keyword. It's the entry point for every workflow.
  • get_serviceFetch full details for one service: rate, min/max quantity and supported options.
  • create_orderPlace an order for a service against a target link or username.
  • order_statusCheck the live status, progress and remaining quantity of an order.
  • balanceRead the current account balance and currency.
  • refillRequest a refill on an eligible order that dropped below its delivered amount.
  • cancelCancel an order that is still cancelable and recover the remaining funds.
  • list_ordersPage through recent orders with their current status.
  • add_fundsCreate a top-up intent to add balance to the account.

The "search-then-act" design

A typical SMM panel carries thousands of services across dozens of platforms. Registering one tool per service would flood the agent's tool list, blow up the context window and make the model slower and less accurate. SocialGO avoids this entirely.

Instead, the catalog lives behind a single discovery tool. The agent first calls search_services to narrow thousands of options down to a handful of relevant matches, inspects one with get_service, and only then acts with create_order. The same nine tools cover the entire catalog no matter how large it grows, so adding new services on the backend never changes the tool surface the agent sees.

A complete flow looks like this:

1. search_services({ query: "instagram followers" })
   → returns a short list of matching services with ids + rates

2. get_service({ id: "<id>" })
   → confirms rate, min/max and options before spending

3. create_order({ id: "<id>", link: "<target>", quantity: 1000 })
   → returns an order id

4. order_status({ id: "<order_id>" })
   → poll until completed

This keeps the tool list tiny and predictable, lets the catalog scale to thousands of services, and gives the agent a reliable confirm-before-spend step every time.