The Model Context Protocol (MCP) is a standard way for AI tools such as Claude, Copilot, or Codex to receive context about a specific business or system. Without it, every AI tool that wants to look at your ecommerce data needs a custom integration for each data source.

Noibu's AI plugin is our implementation of this protocol, acting as a bridge for AI tool access to your Noibu sessions, journeys, errors, performance, and heatmaps data, alongside other components of your ecommerce stack. Once installed, you can ask your AI tool to investigate errors, analyze performance, and find conversion opportunities.

Our plugin also includes pre-built skills to handle some of the most common ecommerce workflows: analyze checkout drop-off, find the root cause of errors, and even draft a code fix for your team to review and merge.

This guide walks through everything needed to set up the Noibu AI plugin for your store.

Using Noibu in AI tools

How far the Noibu plugin can take you depends on what else you connect.

➡ Plain-language queries

On its own, Noibu's AI plugin can help you perform deep analysis across your ecommerce data just by asking in simple language. No dashboards to learn, no exports to stitch together.

➡ Out-of-the-box analyses

Where the AI tool supports them, the plugin also includes pre-built skills for common ecommerce analysis: store health checks, ranked revenue opportunities, checkout funnel analysis, and segment and product deep-dives.

➡ Automated workflows

The biggest unlock comes from connecting the rest of your stack. Connecting a tool like GitHub enables your AI tool to run autonomous, end-to-end workflows — like tracing an error to its root cause and drafting the actual code fix as a pull request, ready for your team to review and merge.

Setting up the plugin

Four steps get the plugin working the way it's designed to. Installing it is enough to start asking questions, but steps 3 and 4 are what make the answers precise and let your AI tool act on what it finds.

Step 1 — Confirm your Noibu access

Your Noibu team provisions your login as part of onboarding, so there is nothing to configure here. Confirm you have your credentials before continuing. If you don't, contact your Noibu team.

Step 2 — Install the plugin / MCP connector for your AI tool

Follow the install guide for the tool your team uses. Each guide covers its own prerequisites, including the plan or licence the AI tool requires. What each tool supports differs, so check the table before choosing where to install the plugin.

Step 3 — Set up ecommerce events for your domain

Ecommerce event data records what shoppers do on your site — adding an item to the cart, starting checkout, placing an order. With them in place, the plugin can connect session and page data to actual purchase behaviour.

• If Noibu was installed through the Shopify app, the BigCommerce app, or the Salesforce Commerce Cloud SFRA cartridge, ecommerce events are already configured.

• For merchants on other platforms, ecommerce event tracking is implemented manually.

Step 4 — Connect GitHub or another CI/CD tool

Connecting a continuous integration and continuous delivery (CI/CD) tool is what turns a diagnosis into a fix. Without it, your AI tool can tell you what's wrong and why; with it, the same investigation can end in a pull request your team reviews.

For GitHub, follow the guide that matches your setup:

• Connecting GitHub to Claude

• Connecting GitHub to OpenAI Codex

• Connecting GitHub to Shopify

Other CI/CD tools usually need a custom implementation. Your Noibu team can work through the right setup for your stack — reach out to get started.

Connecting the rest of your stack

These connections are optional, but recommended. Each one widens what your AI tool can investigate and act on.

Connecting your ecommerce platform

Connecting your platform directly to your AI tool lets it read and act on live store data — products, orders, and theme code — alongside your Noibu data.

• Shopify has a native connector for both Claude and Codex, but requires custom MCP setup for Microsoft Copilot.

• Other platforms usually need a custom implementation with your platform provider. Contact your Noibu team if you want to explore this.

Connecting other third-party tools

The more of your workflow your AI tool can reach, the more of an investigation it can carry end to end:

• Communication tools such as Gmail and Slack — send reports and share findings where your team already works

• Workflow tools such as monday․com — create tasks and update their status as part of an investigation

• Ticketing tools such as Jira — open tickets for issues that need a person to act on them

• Customer support tools such as Gorgias — cross-reference support tickets against technical issues to confirm customer impact, then close the loop with the shoppers who reported them

• Marketing tools such as Google Ads, Meta, and Klaviyo — add campaign context to investigations, so a traffic or conversion shift can be attributed to a campaign rather than a bug

• Cloud and infrastructure tools such as Azure — bring server-side context into root-cause analysis