Canix MCP Server - Setup & Getting Started Guide

Table of Contents

• Overview
• What You Can Do
• What You Need
• Setting Up with Claude
• Setting Up with ChatGPT
• Using the Integration
• Example Questions to try
• FAQs
• Troubleshooting

Overview

The Canix MCP Server lets you connect Canix to AI tools like Claude and ChatGPT using the Model Context Protocol (MCP). Once connected, you can ask natural language questions about your Canix data and get instant answers — no exports or manual lookups required.

Beta Release: This initial release covers Sales Order reporting. Inventory, production, and additional data types are coming in future releases.

Example response:

What You Can Do

• Ask questions like "Show me my top-selling products this month"
• Query sales order data across your facilities
• Get instant summaries and breakdowns without leaving your AI tool

What You Need

• A Canix account
• Claude (desktop or web app) or ChatGPT

Setting Up with Claude

Step-by-Step Instructions

1. Open Claude and go to Settings — click your profile icon in the top-right corner and select Settings.
2. Navigate to the Custom section — in the left sidebar, click on Custom.
3. Click "+ Add" in the top right — then select Add custom connector.
4. Enter a name for the connector — we recommend "Canix" so it's easy to identify.
5. Enter the Canix MCP server URL:

   https://mcp.canix.com/mcp

6. Click Add to save — Canix will now appear in your connectors list.

Tip: You'll be prompted to authenticate with your Canix account the first time you use the connector in a conversation.

Setting Up with ChatGPT

Step 1: Enable Developer Mode

ChatGPT requires Developer Mode to be enabled before you can add custom MCP apps.

1. Open ChatGPT and go to Settings
2. Click on Apps in the left sidebar
3. Click Advanced settings
4. Toggle on Developer Mode (note: OpenAI marks this as Elevated Risk, as it allows unverified connectors)

Step 2: Create the Canix App

1. Go back to Settings → Apps
2. Click Create app
3. Enter a name (e.g. "Canix")
4. Set the Server URL to:

   https://mcp.canix.com/mcp

5. Set Authentication to OAuth
6. Check the acknowledgment box and click Create

Note: OpenAI will display a warning that custom MCP servers introduce risk, as they have not been reviewed by OpenAI. Check the acknowledgment box to proceed.

Using the Integration

Once connected, you can start querying your Canix data directly in conversation.

Example Questions to Try

• "What were my total sales last month?"
• "Show me my top 10 products by revenue this quarter."
• "Which customers placed the most orders in the last 30 days?"
• "Break down sales by facility for this week."
• "What products did I sell most of yesterday?"

Tips for Best Results

• Be specific about time ranges (e.g. "this month", "last 30 days", "Q1 2026")
• Mention your facility name if you have multiple locations
• Ask follow-up questions — the AI remembers context within a conversation

Frequently Asked Questions

Will the server maintain memory of prior conversations?

No — the MCP server itself stores no memory. Memory is handled entirely by the AI assistant you're using (Claude, ChatGPT, Gemini, etc.) and only persists within a single conversation. If Canix introduces a native chat UI in the future, we'll add a memory mechanism at that point.

Will data be limited to my currently selected licenses, or all licenses in my environment?

You'll have access to data from all facilities your Canix account has permission to view. You can narrow the scope at any time just by mentioning it in your message — for example: "Generate a sales report just for my distribution facilities."

Can the model present data across all Canix-integrated licenses at the state or national level?

Yes — if your account has access to multiple licenses across locations, you can query and report across all of them.

Will the MCP server's answers respect my user permissions?

Yes — the MCP server enforces Canix's existing user access controls. You will only be able to query data that your account has permission to view.

Troubleshooting