AI Integration
Unchained Engine provides first-class AI integration through three complementary surfaces:
| Surface | Who uses it | Protocol | Auth | Endpoint |
|---|---|---|---|---|
| MCP Server | AI agents & IDEs (Claude Desktop, Cursor, Claude Code) | Streamable HTTP (MCP) | Admin bearer token | /mcp on the engine |
| Admin Copilot | Store operators via Admin UI chat | HTTP streaming (Vercel AI SDK) | Session cookie | /chat on the engine |
llms.txt | LLMs & web crawlers | Static files | None | docs.unchained.shop/llms.txt |
Architecture
The MCP server exposes the full commerce API as AI-callable tools. The Chat API (used by the Admin UI Copilot) connects to the MCP server internally and streams responses back to the browser.
Quick links
- MCP Server Reference — Tool categories, resources, authentication, and connection examples for AI agents
- Admin Copilot Setup — Configure the built-in chat assistant in the Admin UI
- AI FAQ — Common questions about AI capabilities, answered for both humans and AI agents
llms.txt
This documentation site (not the engine) publishes /llms.txt and /llms-full.txt following the llms.txt standard. These files help AI models discover and navigate the documentation efficiently.