We asked Claude to plan a Black Friday campaign.
The AI agent analysed last year's revenue, proposed six staggered campaigns across Nordic markets, and set up every single one of them directly through Brink Commerce APIs. It took about twenty minutes of skilled operator work and compressed it into a single, three-minute conversation. No code, no manual clicking through menus, and every action was cleanly recorded in the portal's audit log (read on to see it in action).
That is the MCP for Merchant Portal, a new capability rolling out to our merchants. It allows your operations teams to manage your entire commerce engine using natural language or autonomous agents. By connecting Claude, ChatGPT, or your own in-house AI framework, you can search orders, handle refunds, configure markets, adjust inventory rules, and manage campaigns without navigating a traditional admin interface.
Anyone running day-to-day eCommerce operations knows the familiar pattern. A simple customer inquiry like "why was this discount rejected at checkout?" kicks off a multi-step tour. You find yourself clicking from the rule editor to the order list, jumping over to the sales dashboard, and perhaps even running an ad-hoc API call to see what went wrong.
Each individual task is small, but the constant jumping between views is where your team's time actually goes. As an enterprise grows, this complexity multiplies. Every new market, currency, warehouse, and promotion rule adds more views to navigate and more potential failure points.
The MCP for Merchant Portal removes this friction. Instead of forcing your team to navigate to the work, it enables them to describe the work and let their agent execute it.
MCP stands for Model Context Protocol. Developed and open-sourced by Anthropic, and has since been adopted by OpenAI and Google, it functions like a universal USB port connecting AI models to secure data and real-world systems.
Because we built this on an open standard, there is no proprietary platform to adopt. If your business already uses Claude, ChatGPT, or a custom internal framework, you can connect it directly to Brink Commerce over HTTP and start working immediately.
This agentic interface works because Brink is built from the ground up as a headless, API-only system. Because all platform services are exposed as clean, modular endpoints, an AI agent can read, validate, and execute operations just like a human developer would.
"We did not want to build another closed-loop AI assistant," says John Osvald, CPO at Brink Commerce. "Merchants are already building their own internal tooling and choosing their preferred AI partners. By exposing Brink's Management and OMS APIs as typed, schema-defined tools, any standard-compliant agent can understand our system immediately. The agent does not have to guess payloads from documentation: it executes real, validated operations directly."
It is important to clarify what this is not. This is not a generic chatbot bolted onto a dashboard to summarise text. It is a highly secure, controlled execution plane.
"This is the part that standard AI demos usually skip," John explains. "You cannot let an autonomous agent roam free in an enterprise commerce database without guardrails. So we designed MCP to respect the exact same security architecture that governs our platform."
To show how this works in practice, we recorded two live sessions running against real Brink Commerce APIs over MCP. Together, they demonstrate the classic operational loop: observe what is happening, decide on a course of action, and execute.
In the video at the top of this post, we tasked Claude with preparing our holiday promotion strategy. After analysing historical performance, the agent drafted a detailed, multi-market campaign plan. Once approved, the agent called the Brink Commerce Management API to construct the rules, markets, and dates automatically. What used to be a long configuration session is reduced to an audit review of a finished setup.
In our second walkthrough, we focus on diagnostics and insights. By querying our shopper insights index, the operator asks the agent to analyse week-over-week activity.
[VIDEO: Finding conversion drops with Shopper Insights, 40s]
The agent runs live aggregations, identifies a fifty percent drop in activity, and isolates three products that were added to carts but abandoned. Crucially, it notes that two of those items are low on stock, and suggests a targeted, time-limited recovery campaign to run alongside our summer sale.
Because diagnostic tools and execution systems share the same interface, resolving the issue is as simple as replying: "Go ahead and set up that recovery promotion."
The MCP for Merchant Portal covers the full scope of your daily administrative tasks. Operations, customer service, and eCommerce teams can manage several key areas:
Discount management is inherently complex. Rules often conflict across markets, and small configuration errors can have immediate revenue impacts. Consequently, discounts are one of the most common sources of daily support inquiries.
MCP simplifies this complexity by condensing multi-step setups, testing, and diagnostics into direct statements:
The immediate win is operational speed. Complex, repetitive workflows that used to require clicking through four different screens can now be completed in seconds. Issues are resolved faster, and operations teams spend less time on manual admin.
The strategic win is the path it creates toward automation.
Because MCP uses a single, open interface for both human instructions and machine inputs, it provides a natural path from assisted tasks to complete automation. A workflow that your team runs by hand today can easily be transitioned into a scheduled background script next month, and eventually into a fully autonomous agent.
You can make this transition gradually and safely, without rewriting code or changing your underlying API integrations. You gain the freedom to build your own operational assistants directly on top of Brink Commerce, keeping your technology stack flexible and ready for whatever comes next.
MCP access is rolling out to all Brink Commerce merchants. It can be enabled per user through a dedicated Merchant Portal role, allowing you to run testing in a staged environment before deploying to production.
The traditional Merchant Portal UI remains entirely unchanged.
