Skip to main content

Setup & Configuration

Install & Build

1

Install dependencies

2

Build the server

This compiles TypeScript to dist/index.js — the entry point your MCP client will run.
3

Set environment variables

Create a .env file in the zeplo-mcp-server directory (or set them in your client config):
The server needs read/write access to the same PostgreSQL database used by your Zeplo instance.

Configure Your Client

Add to your project’s .claude/settings.json:
Or via the CLI:

Verify It Works

Restart your AI assistant, then ask it to check system health:
If configured correctly, the assistant will call the system_health tool and report the database connection status.

Environment Variables

Development

For local development without building each time:
The MCP server communicates over stdio — it reads from stdin and writes to stdout. Logs and errors go to stderr. Your MCP client handles the transport automatically.

Dry-Run Mode

Tools that create or modify data (contact_create, campaign_create, message_send_text, message_send_template, chatbot_train) support a dryRun parameter. Set dryRun: true to validate inputs without writing to the database.

Error Handling

The server returns structured errors for:
  • Not found — resource with the given ID doesn’t exist
  • Validation failed — input doesn’t match the expected schema
  • Database error — connection issues or constraint violations
  • Automation not active — attempting to execute a paused or inactive automation
All errors include a descriptive message explaining what went wrong.