Model Context Protocol

CRM Lead MCP Server

Connect Codex or Claude to create one lead or securely process a CSV or Excel file with row-level results.

Open hosted chatConfigure MCP client
Health
/api/health
MCP endpoint
/api/mcp
Transport
Streamable HTTP

Customer setup

Connect your AI client

Pick your operating system first, then configure the AI client you use. Restart the client after saving environment variables.

3 steps
  1. 1

    Save your CRM credentials

    These variables stay on your machine. The MCP client sends them as request headers so the server can retrieve a CRM token for your tenant.

    Windows - PowerShell

    Run in a new PowerShell window, then completely restart Codex or Claude.

    setx CRM_CLIENT_ID "your-client-id"
setx CRM_CLIENT_SECRET "your-client-secret"
  2. 2

    Configure your AI client

    Choose the client your customer uses. Codex reads credentials from environment variables; Claude Code can receive headers directly in the MCP command.

    Codex config.toml

    Add this to your Codex config file, then restart Codex. On Windows this file is usually %USERPROFILE%\.codex\config.toml. On macOS and Linux it is usually ~/.codex/config.toml.

    [mcp_servers.mca_mcp]
enabled = true
url = "https://crm3qamcp.invogen.org/api/mcp"
env_http_headers = { "X-CRM-Client-ID" = "CRM_CLIENT_ID", "X-CRM-Client-Secret" = "CRM_CLIENT_SECRET" }
  3. 3

    Test, then create leads

    Start a new conversation and use these prompts:

    Use mca_mcp to test CRM authentication.Use mca_mcp to create a lead for John Doe at Acme Corporation. Use Acme Corporation as the DBA.Use mca_mcp to process C:\Data\leads.xlsx and return the status for every row.

Keep your client secret private. Use this service over HTTPS outside local development.