AI Integration
Make your Juice apps discoverable and assistable by AI coding tools. Juice ships two features for this: an llms.txt file for LLM discovery and an MCP server for deep framework knowledge.
Why This Matters
AI coding assistants (Claude, Cursor, Copilot) work best when they understand your framework. Without explicit context, they hallucinate patterns from other frameworks — getServerSideProps instead of async server components, next/link instead of Link from @cmj/juice/client.
Juice solves this at two levels:
- llms.txt — A standard text file at your docs root that any LLM can read to understand your framework. Zero setup, works everywhere.
- MCP Server — A Model Context Protocol server that exposes searchable docs, API reference, concept explanations, and code scaffolding as tools. Deep integration for AI editors that support MCP.
Quick Comparison
| llms.txt | MCP Server | |
|---|---|---|
| Setup | None — served at /llms.txt | Add to MCP config |
| How it works | LLM fetches and reads a single file | LLM calls tools for specific queries |
| Depth | Overview + links | Full docs, API lookup, code generation |
| Best for | Any LLM, any tool | Claude Desktop, Claude Code, MCP-enabled editors |
Getting Started
If your AI tool supports MCP, start with the MCP Server — it gives the richest experience. Otherwise, point your tool at the llms.txt URL and it will have everything it needs.