Skip to main content
This quickstart gives you both integration paths:
  1. Direct MCP for clients that support streamable_http
  2. HoopAI MCP Proxy for clients that only support stdio

Prerequisites

  • A HoopAI Private Integration Token (PIT) or OAuth access token
  • A locationId to scope tool calls
  • Scopes for the modules you plan to use (contacts, conversations, calendars, opportunities, payments, etc.)
Use this when your MCP client supports HTTP Streamable transport. Endpoint
https://services.leadconnectorhq.com/mcp/
Required headers
  • Authorization: Bearer <token>
  • locationId: <your-location-id>

Example: LangGraph MCP client

from langchain_mcp_adapters.client import MultiServerMCPClient

client = MultiServerMCPClient(
    {
        "ghl": {
            "url": "https://services.leadconnectorhq.com/mcp/",
            "transport": "streamable_http",
            "headers": {
                "Authorization": "Bearer your-token",
                "locationId": "your-location-id"
            }
        }
    }
)

Example: n8n MCP client

  • MCP URL: https://services.leadconnectorhq.com/mcp/
  • Transport: HTTP Streamable
  • Header: Authorization with Bearer <token>
  • Provide locationId as header or prompt context
The official MCP docs call out HTTP Streamable support in n8n v1.104+.

Option B: HoopAI MCP Proxy (stdio)

Use this path for MCP clients that do not yet support HTTP Streamable (for example Claude Desktop).
{
  "mcpServers": {
    "hoopai": {
      "command": "npx",
      "args": [
        "@hoopai/leadconnector-mcp-proxy"
      ],
      "env": {
        "HOOPAI_ACCESS_TOKEN": "your-token",
        "HOOPAI_LOCATION_ID": "your-location-id",
        "HOOPAI_MCP_SERVER_URL": "https://services.leadconnectorhq.com/mcp/",
        "HOOPAI_MCP_MODE": "header"
      }
    }
  }
}

Smoke test prompts

After setup, test with prompts like:
  • “Show me my last 10 contacts”
  • “Find unread conversations and summarize them”
  • “Check next available slots for tomorrow”
  • “Move opportunity X to the next pipeline stage”

Troubleshooting

  • 401/403: token is invalid or missing scopes.
  • 400: wrong or missing locationId.
  • No tool responses in stdio clients: use the HoopAI proxy path.
  • Wrong location data: explicitly set HOOPAI_LOCATION_ID.
Last modified on March 6, 2026