Building an AI-powered chat with Strapi and MCP that brings customers and staff closer together

A developer describes an open-source prototype for a small luxury retailer that combines LINE messaging, Strapi 5, MCP tools, TanStack Start, and TanStack AI to provide customer support, appointment requests, and staff oversight. The system centralizes product data, inventory, store details, booking rules, and bilingual knowledge in Strapi, while exposing 13 permission-controlled MCP tools that both the customer app and AI agents use under the same business rules. Customers authenticate through LINE, allowing tools to derive identity from the session rather than model-provided inputs, and the concierge can answer only from approved knowledge, resolve dates through dedicated tools, initiate a customer-confirmed booking form, or automatically hand unanswered questions to staff. Appointment requests remain drafts until staff approve them, after which confirmations are sent through LINE. Staff work entirely within a Strapi admin extension featuring request, question, and inquiry queues, dashboards, and a read-only AI assistant whose available tools depend on role permissions. Conversations are logged and automatically labeled for priority, while staff responses can be saved as new knowledge, gradually improving future concierge answers. The author emphasizes task-oriented tool design, session-based identity, human approval for irreversible actions, code-enforced safeguards, and fixed tool lists, and provides repository paths and prompts for adapting the architecture to other businesses.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.