A practical guide to the Agentic Protocol Shopify integration

“Agentic protocol Shopify integration” can describe several different projects. A store owner might mean making products available in AI shopping channels. A developer might mean building a shopping agent. A support lead might mean connecting an AI teammate to Shopify order data.

Those projects can complement each other, but they are not the same integration. This guide separates them and shows how eesel CLI helps with the support work after a customer places an order.

Start by identifying which integration you need

Connecting the support teammate does not enroll a store in a shopping channel. Likewise, being discoverable in an AI channel does not configure your support escalation rules.

Shopify agentic storefronts: the merchant path

Shopify’s current help documentation says agentic storefronts are active by default for eligible stores. Merchants manage their AI channels under Sales channels > Agentic in Shopify admin.

Checkout depends on the channel. The documentation describes ChatGPT as a discovery referrer: customers finish on the merchant’s checkout through an in-app browser or a new tab. Other listed channels can use Shopify-powered direct checkout when activated.

Orders retain channel or referrer attribution in Shopify admin, and the merchant remains responsible for the post-purchase experience. Check your store’s eligibility, terms, and actual channel settings rather than assuming every AI referral follows the same checkout.

UCP and MCP: the developer path

Shopify’s developer overview now centers on Universal Commerce Protocol, or UCP, and Shopify’s UCP-compatible MCP servers.

UCP describes commerce interactions. MCP exposes tools an agent can call. They are not interchangeable names, and MCP is not a protocol owned by Shopify.

The developer offering spans product discovery, carts, checkout, and order tracking. Access and trust levels affect what an agent may do. Shopify also lists a UCP CLI and toolkit for building those experiences. You can request early access to Universal Cart.

ACP, the Agentic Commerce Protocol developed by Stripe and OpenAI, is a separate commerce standard. Choose a path based on the channel and integration you are implementing.

Why the support workflow is still separate

Suppose a customer discovers a product through an AI assistant, buys it, and later asks why the delivery estimate changed. Your support team needs the order, current fulfillment information, and the applicable shipping policy.

The referral source alone cannot answer that question. Neither can the shopping agent’s recommendation. Be clear about which system holds each fact and what a teammate should do when it lacks evidence.

Before automating this case, decide:

• How the customer is verified before private order information is shared.
• Which source establishes the current order or shipment state.
• Which policy governs the promised next step.
• When a human must review an exception or order change.

Those decisions are useful whether the purchase began in an AI conversation, a search engine, or your storefront.

Connect and inspect the support teammate through eesel CLI

eesel CLI operates the same eesel agent and workspace used in the dashboard. It gives a person terminal access, lets scripts query operational information, and returns JSON that coding agents such as Claude Code, Cursor, and Codex can read.

For Shopify support, the value is concrete: inspect the teammate’s store access, check its instructions, and review what it did. You are operating your support teammate—not Shopify’s UCP CLI or the agentic sales-channel settings.

1. Identify the teammate before adding access

With Node.js 18.17 or newer:

npx @eesel/cli login
npx @eesel/cli whoami
npx @eesel/cli agents

Replace AGENT_ID with the intended teammate’s ID:

npx @eesel/cli integrations --agent AGENT_ID
npx @eesel/cli instructions --agent AGENT_ID
npx @eesel/cli automations --agent AGENT_ID

Review existing automations and permitted actions before connecting the store. An enabled workflow may already be able to act when more data or actions become available. A new CLI session does not mean a new, inactive teammate.

2. Choose the right Shopify connection

The eesel Shopify guide distinguishes a public storefront crawl from the full Shopify connection.

A crawl adds publicly visible storefront knowledge. The full connection requires a person to approve eesel in Shopify admin and provides additional synced sources and permitted live lookups. It is the relevant path when support needs private order information.

You can begin through chat, after reviewing action permissions:

npx @eesel/cli chat "Help me connect my Shopify store." --agent AGENT_ID

Chat can invoke actions, so this is a setup request, not a read-only inspection. Provide the requested store name and have an authorized person review and complete the approval in the browser.

3. Check synced knowledge and live data separately

Shopify product and policy sources are synced. Lookup actions query Shopify live. A downloaded shipping-policy page does not establish the current location of a parcel.

Use status to inspect operational readiness:

npx @eesel/cli status --agent AGENT_ID

If you add an approved local support policy:

npx @eesel/cli files upload ./delivery-exceptions.pdf --agent AGENT_ID

Replace that example path with your document. The upload adds knowledge; it does not rewrite standing instructions or grant permission to alter orders. Review those separately if the policy changes who can authorize an exception.

Status and source counts are not quality scores. Test a normal delivery question, missing order details, and a case that needs escalation. Check the evidence behind each answer.

4. Start with restricted actions, then inspect the work

Shopify supplies sources and actions to eesel, but no triggers. A helpdesk conversation or storefront chat starts the workflow. Configure when the teammate should use Shopify and keep consequential actions off or subject to approval while testing.

Inspect its activity and held actions:

npx @eesel/cli activity --agent AGENT_ID
npx @eesel/cli approvals --agent AGENT_ID

Compare the response with the actual support conversation and order record. Did it find the intended order? Did it avoid promising a delivery date without evidence? Did it hand off the exception?

Creating or enabling automation is a separate step from checking readiness. Review scope and action settings before allowing customer-facing work.

Give a coding agent a specific review task

For example:

Inspect this eesel teammate’s Shopify connection, instructions, and automations. Report missing sources, unclear delivery-exception rules, and actions that can change orders. Ask before making changes.

JSON output makes that task accessible to a coding agent. The support lead can review the same teammate in the dashboard. Asking first is an instruction to the coding agent, not a replacement for eesel’s configured action approvals.

For scripts, select the agent explicitly, keep EESEL_API_TOKEN secret, and verify EESEL_API_URL points to the intended server. Use --help for command options. A write command’s --dry-run prints the server request without sending it; it does not simulate a support conversation or test a checkout.

Prepare each part of the customer journey

Use Shopify admin to review shopping-channel participation, and current Shopify developer documentation if you are building a commerce agent. Treat support access and permissions as their own setup.

Try eesel and use eesel CLI to inspect one Shopify support workflow. Start with a narrow case, review the answer and actual actions, and expand once the evidence supports it. That prepares your team for the customer’s problem, regardless of where the purchase began.