Package Exports

@metaengine/mcp-server
@metaengine/mcp-server/dist/index.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@metaengine/mcp-server) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

MetaEngine MCP Server

Turn one conversation into 50 consistent files. Claude architects, MetaEngine builds.

A specialized tool for AI assistants. You describe what you need in natural language—Claude decides when structured generation makes sense and handles the rest.

Install

Add to ~/.claude/mcp.json (Claude Code) or Claude Desktop config:

{
  "mcpServers": {
    "metaengine": {
      "command": "npx",
      "args": ["-y", "@metaengine/mcp-server"]
    }
  }
}

That's it. No API key, no signup, free to use.

How It Works

Without MCP (file-by-file):

Claude: Generate User.ts → write → generate Order.ts → write → generate Product.ts → write...
Result: 20 sequential operations, growing context, potential pattern drift

With MCP (structured):

You: "Create User, Order, and Product models with services and repositories"
Claude: Constructs spec → MetaEngine returns all files → writes 15 files at once
Result: Instant, consistent, minimal context usage

You talk to Claude in natural language. Claude decides when structured generation makes sense and handles everything automatically.

When Claude Uses This Tool

Generating 20-100 interconnected files where consistency matters
Multi-language projects (same architecture in TypeScript, Python, Go, C#, Java, Kotlin, Groovy, Scala, Swift, PHP)
Pattern multiplication (same structure applied to many entities)
Complex import management across deep namespaces

When Claude Generates Directly

1-15 files — Claude's direct generation is faster for small tasks
Exploratory coding where structure is still evolving
One-off scripts and utilities
Rapid prototyping where flexibility beats consistency

Both approaches are valid. Claude picks the right tool for the job.

Real-World Workflows

Cross-Language Test Parity

Point Claude at existing TypeScript tests → generate equivalent Go/Python/C# tests with same coverage in minutes instead of days.

Multi-Entity CRUD System

"Generate User, Order, Product, Category with services, repositories, DTOs, and validators" → 50 files in ~60 seconds with perfect consistency.

Polyglot Architecture

Same domain model across TypeScript frontend, Python API, Go microservices. One conversation, consistent types everywhere.

Pattern Multiplication

Claude reads one well-designed service, generates the same pattern for 20 entities with identical structure.

Performance

Instant generation — 25-100 files in one call
3.8x faster than file-by-file generation at scale
57% fewer tokens — compact spec vs full file text

The tipping point is around 20 files—below that, Claude's direct generation is simpler.

Languages Supported

TypeScript, Python, Go, C#, Java, Kotlin, Groovy, Scala, Swift, PHP

Each generates idiomatic code (data classes in Kotlin, case classes in Scala, structs in Swift, etc.)

Privacy & Pricing

Free — no API key, no signup, unlimited requests
Private — specs sent for generation are never saved or logged
Local — MCP server runs on your machine, MIT licensed

Documentation

The AI guide is automatically embedded in the tool description on first use — no manual reading required. After the first successful call, the description swaps to a lightweight directive to minimize context usage.

METAENGINE_AI_GUIDE.md — Critical rules, patterns, language notes, and common mistakes
EXAMPLES.md — Real-world usage with input/output across all languages

Common Questions

Do I need to learn the JSON spec format? No. You talk to Claude in natural language. Claude constructs specs automatically.

How is this different from Claude generating code directly? Claude Code is excellent at direct generation—and faster for small tasks. MetaEngine is a specialized tool for when you need structured generation at scale (20-100 files) with guaranteed consistency.

Is my code sent to a server? Only the generation specs (JSON describing what to create). Your existing code never leaves your machine. Specs are never saved or logged.

Can I use this with other AI assistants? Yes—any MCP-compatible assistant works. Currently that includes Claude Desktop and Claude Code.

License

MIT