CodeApp SDK

Use Claude Code as an LLM provider with your Claude Code subscription flat fee instead of pay-per-token API keys.

Installation

npm install codeapp_code_sdk

Quick Start

1. Create a new project

npx codeapp-sdk init my-app
cd my-app
npm install
npm start

2. Open in browser

Visit http://localhost:3000 to see the chat interface.

3. Test the API

curl -X POST http://localhost:3000/chat \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Hello Claude!"}'

Features

• 💰 Flat Fee: Use your Claude Code Max subscription for unlimited LLM usage
• 🚀 No Token Counting: Stop worrying about API costs
• 🔌 Simple Integration: Drop-in SDK for any JavaScript/TypeScript project
• 🌐 WebSocket Support: Built-in Socket.io integration for real-time applications
• 📦 CLI Tool: Quick project scaffolding with codeapp-sdk init
• 🔄 Streaming Support: Real-time response streaming with AsyncGenerator

Usage

Basic SDK Usage

import { ClaudeCodeSDK } from 'codeapp_code_sdk'

const sdk = new ClaudeCodeSDK()

// Simple prompt
const response = await sdk.executeWithClaudeCode('Hello, Claude!')
console.log(response.messages[0].content)

// With options
const response = await sdk.executeWithClaudeCode('Explain quantum computing', {
  modelId: 'sonnet',     // or 'opus', 'haiku'
  maxTokens: 1000,
  temperature: 0.7
})

Streaming Responses

const stream = sdk.streamWithClaudeCode('Write a story...')
for await (const chunk of stream) {
  process.stdout.write(chunk)
}

WebSocket Server

import { ClaudeCodeWebSocketServer } from 'codeapp_code_sdk'

const server = new ClaudeCodeWebSocketServer({ port: 8080 })
// Connect with ws://localhost:8080

Socket.io Integration

import { createServer } from 'http'
import { ClaudeCodeSocketIOServer } from 'codeapp_code_sdk'

const httpServer = createServer()
const claudeServer = new ClaudeCodeSocketIOServer(httpServer)

httpServer.listen(3000)

API Reference

ClaudeCodeSDK

new ClaudeCodeSDK(options?)

• options.debug: Enable debug logging (default: false)

executeWithClaudeCode(prompt, options?)

• prompt: The input prompt string
• options.modelId: Model to use ('sonnet', 'opus', 'haiku')
• options.maxTokens: Maximum tokens to generate
• options.temperature: Temperature for generation (0-1)
• options.systemPrompt: System prompt (can include '--continue' for conversation continuity)

Returns: Promise<ClaudeCodeResponse>

streamWithClaudeCode(prompt, options?)

Returns: AsyncGenerator<string> for streaming responses

stopCurrentProcess()

Stops the current Claude Code process if running.

getCurrentConversationId()

Returns the current conversation ID for continuity.

CLI Commands

• codeapp-sdk init [project-name] - Create a new CodeApp SDK project
• codeapp-sdk check - Verify Claude Code installation
• codeapp-sdk serve [--port --token ...] - Start middleware service
• codeapp-sdk help - Show help information

Requirements

• Node.js >= 18.0.0
• Claude Code CLI installed and authenticated
• Active Claude Code subscription

Debugging

Common Issues

1. "Claude Code not found": Make sure Claude Code CLI is installed and in your PATH
2. "Failed to connect": Ensure the server is running (npm start)
3. CORS errors: The server includes CORS support by default
4. Model errors: Not all models may be available with your subscription

Debug Mode

Enable debug logging to see detailed information:

const sdk = new ClaudeCodeSDK({ debug: true })

License

MIT

Contributing

See CONTRIBUTING.md

Links

• GitHub Repository
• npm Package
• Claude Code