AI Speeds

AI content platform and universal Claude API proxy. Translates Anthropic Messages API to OpenAI-compatible format, enabling Claude Code to work with any provider.

Deployed at aispeeds.me.

Features

• API Proxy -- Anthropic Messages API to OpenAI Chat Completions / Responses API with real-time SSE streaming
• Multi-Provider -- DeepSeek, OpenAI, OpenRouter, Kimi, SiliconFlow, NVIDIA NIM, or any OpenAI-compatible endpoint
• React Homepage -- Native Next.js landing and get-started guide for product makers and Claude Code users
• Edge Deploy -- Cloudflare Workers (300+ locations, sub-ms cold start)

---

Application Tabs

The app has 4 main navigation tabs, each serving a distinct purpose:

1. Home -- Product Maker Homepage

Route: /

The default landing page is a native React/Tailwind experience for AI product makers. It introduces AI Speeds, Claude Code Router, and quick entry points for the get-started guide, whiteboard, and API playground.

2. Get Started -- How to Use Claude Code

Route: /home#get-started

Step-by-step guide for setting up Claude Code with various AI providers: install Claude Code, choose a provider, and configure environment variables.

3. Whiteboard

Route: /whiteboard

Free-form drawing and writing tool powered by Excalidraw. Useful for quick sketches, wireframes, and visual brainstorming.

4. Playground -- API Testing

Route: /playground

Built-in API test tool supporting three API formats:

• OpenAI Chat Completions (/v1/chat/completions)
• OpenAI Responses (/v1/responses)
• Anthropic Messages (/v1/messages)

Features streaming output (rendered + raw SSE view), model discovery, and latency measurement.

---

Quick Start

Use the hosted instance

# Install Claude Code
pnpm add -g @anthropic-ai/claude-code

# Point it at AI Speeds
export ANTHROPIC_BASE_URL="https://aispeeds.me"
export ANTHROPIC_API_KEY="your-provider-api-key"

# Start using
claude

Self-host

git clone https://github.com/your-username/ai-speeds
cd ai-speeds
pnpm install
pnpm run cf:deploy

Set environment variables in Cloudflare dashboard or via Wrangler:

Variable	Purpose
DEEPSEEK_BASE_URL	DeepSeek API backend
OPENAI_BASE_URL	OpenAI API backend
KIMI_BASE_URL	Kimi API backend
SILICONFLOW_BASE_URL	SiliconFlow API backend
OPENROUTER_BASE_URL	OpenRouter API backend
NVIDIA_NIM_BASE_URL	NVIDIA NIM API backend
OPENAI_COMPATIBLE_BASE_URL	Generic fallback (auto-detects provider)

Provider selection is automatic based on which *_BASE_URL env vars are set, in the order listed above.

---

Architecture

POST /api/v1/messages  (Anthropic format)
        |
        v
  selectProvider()  -- picks backend from env vars
        |
        v
  formatAnthropicToOpenAI()  -- converts request format
        |
        v
  POST {provider}/chat/completions  (OpenAI format)
        |
        v
  streamOpenAIToAnthropic()  -- converts response back
        |
        v
  SSE stream  (Anthropic format)

Project Structure

src/
  app/
    page.tsx                     # Home page shell with hash-based sections
    api/
      v1/messages/route.ts       # Claude API proxy endpoint
      playground/route.ts        # API test proxy
      playground/models/route.ts # Model list fetcher
      img-proxy/route.ts         # CORS image proxy
    (main)/
      home/page.tsx              # /home route for get-started anchor links
      playground/page.tsx        # API playground UI
      whiteboard/page.tsx        # Whiteboard (Excalidraw)
      brand/page.tsx             # Brand kit
  components/
    HomePageWithNav.tsx           # React homepage, navigation, and get-started guide
    features/whiteboard/          # Whiteboard components
  services/
    llm-provider/
      providers.ts                # Provider configs & model mappings
      adapters/
        format.ts                 # Anthropic <-> OpenAI message conversion
        stream.ts                 # OpenAI SSE -> Anthropic SSE streaming
  config/
    navigation.ts                 # Navigation section IDs
    providers.ts                  # Provider card data for get-started guide
    ui-texts.ts                   # UI text constants (tab labels, etc.)
  lib/                            # Utilities

Tech Stack

• Framework: Next.js 15 (App Router) + React 19
• Styling: Tailwind CSS 3
• Language: TypeScript (strict mode)
• Runtime: Edge Runtime on Cloudflare Workers via @opennextjs/cloudflare
• Build: Turbopack (dev) + Next.js production build

---

Development

pnpm install
pnpm run dev             # Dev server at localhost:3000

Commands

Command	Description
pnpm run dev	Dev server with Turbopack
pnpm run build	Production Next.js build
pnpm run typecheck	TypeScript type checking
pnpm run lint	ESLint with auto-fix
pnpm run format	Prettier formatting
pnpm run cf:build	Build for Cloudflare Workers
pnpm run cf:preview	Local preview of CF build
pnpm run cf:deploy	Deploy to production

API Usage

Proxy a Claude request

curl -X POST https://aispeeds.me/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: your-provider-api-key" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Hello"}],
    "max_tokens": 100,
    "stream": true
  }'

The proxy accepts Anthropic Messages API format and forwards to the configured OpenAI-compatible backend. Model names are automatically mapped (e.g. claude-sonnet-4-5-20250929 -> provider-specific model ID).

License

MIT

This is an independent tool, not affiliated with Anthropic, OpenAI, or OpenRouter. Users are responsible for compliance with all relevant Terms of Service.