Build a store support MCP server with an OpenRouter copilot and Slack feedback

## Build a store support MCP server and AI copilot that report tool gaps to Slack This template gives you an MCP server for store support (order lookups, return history, and returns that follow the store's rules) and an AI copilot that uses it like any MCP client. The highlight is send_feedback: agents call it when a tool gets in their way, and each report is stored and posted to Slack. Its goal: rather than quietly working around a missing capability or an unclear error, agents tell you which tool to add or fix next. It runs on demo orders. ### How it works 1. Store Support MCP (MCP Server Trigger) exposes five tools: find_orders, get_order and get_returns read the data tables; create_return and send_feedback call this same workflow. 2. Route by Tool runs the matching backend: the return rules (delivered, 30-day window, final sale, quantity left), or feedback validation, rate limiting, scrubbing, storage and a Slack message. 3. Support Copilot (AI Agent) answers support staff through the MCP Client and files feedback when no tool fits. ### Setup 1. Run Set Up Demo Data: it creates the three data tables and seeds the demo orders. 2. Create a Bearer Auth credential with a long random token and select it on Store Support MCP and Store Support. 3. Select your OpenRouter credential on Copilot Model. 4. Optional: select a Slack credential and channel on Notify Slack. Feedback is stored even without Slack. 5. Publish the workflow, open the chat and ask: "Liam wants to swap his jeans on FN-10457 for a 34." ### Customization - Connect Claude, Cursor or any MCP client to the same endpoint. - Replace the data tables with your own order and returns system, and keep send_feedback as is. Built by [](https://low-code-consultant.com)