Flary NPM | npm.io

Flary

A versatile toolkit for building and managing Cloudflare Workers applications, including AI integrations, project scaffolding, and component management.

Prerequisites

Before you begin, ensure you have:

Node.js 18.x or later installed
A Cloudflare account
Wrangler CLI installed and configured:
```
npm install -g wrangler
wrangler login
```

Installation

npm install flary

Features

🚀 Project Scaffolding: Quickly set up modern Cloudflare Workers projects.
🎨 Component Management: Import and organize components from v0.dev.
🤖 AI Integration: Tools and utilities for AI applications.
🔒 Secure MCP Integration: Model Context Protocol integration with robust authentication.

CLI Usage

Project Initialization

Create a new Flary project interactively:

flary init

This command will:

Prompt for your project name.
Allow template selection (e.g., Full-Stack React App).
Set up a complete project structure.
Configure dependencies and Wrangler automatically.

Follow the provided steps after creation:

Navigate to your project:
```
cd your-project-name
```
Install dependencies:
```
npm install
```
Start development server:
```
npm run dev
```

v0.dev Component Import

Import and organize components from v0.dev:

# Default directory (src/components)
flary v0 "https://v0.dev/chat/your-component-url"

# Custom directory
flary v0 --dir game-components "https://v0.dev/chat/your-component-url"

Options:

-d, --dir <directory>: Specify target subdirectory (relative to src/).
-h, --help: Display help information.

Deployment

Deploy your Flary application to Cloudflare Workers:

Build your application:
```
npm run build
```

Deploy to Cloudflare Workers:

npm run deploy
# or directly with wrangler
wrangler deploy

Your application will be deployed to https://<project-name>.<your-subdomain>.workers.dev

Environment Variables

Set environment variables for production:

# Set a secret
wrangler secret put MY_SECRET
# Set a variable in wrangler.toml
wrangler deploy --var MY_VAR=value

MCP Integration

Basic Usage

Initialize MCP instance and register tools:

import { z } from "zod";
import { MCP } from "flary";

const app = new MCP({
  name: "test-mcp",
  description: "A test MCP",
  version: "1.0.0",
});

const sumSchema = z
  .object({
    a: z.number().describe("The first number to add"),
    b: z.number().describe("The second number to add"),
  })
  .describe("Calculate the sum of two numbers");

async function calculateSum({ a, b }: z.input<typeof sumSchema>) {
  return a + b;
}
calculateSum.schema = sumSchema;

app.tool("calculate_sum", calculateSum);

export default app;
export const { McpObject } = app;

Authentication

Supports Bearer token authentication:

const app = new MCP({
  auth: {
    type: "bearer",
    token: "your-secret-token",
    validate: async (token) => token === (await getValidToken()),
  },
});

Connection Types

SSE (Server-Sent Events): Connect via / or /sse.
WebSockets: Connect via / or /ws.

Cursor Integration

Cursor uses SSE exclusively:

https://your-worker.workers.dev/?key=your-secret-token

No Authentication

Publicly accessible MCP server:

const app = new MCP({
  name: "test-mcp",
  description: "A test MCP",
  version: "1.0.0",
});

Durable Objects Setup

MCP requires Durable Objects for state management. Configure them in your wrangler.json:

{
  "durable_objects": {
    "bindings": [{ "name": "McpObject", "class_name": "McpObject" }]
  },
  "migrations": [{ "tag": "v1", "new_classes": ["McpObject"] }]
}

Make sure to export the McpObject from your worker entry point as shown in the Basic Usage example.