kakitiro.com

AIProxyServer - Top

AIProxyServer is a local proxy that exposes a single OpenAI-compatible API for eleven major cloud AI providers. Your other apps connect to http://localhost, while AIProxyServer holds the API keys, rotates the upstream requests, and translates each provider's protocol into the OpenAI format you already know.

AIProxyServer

AIProxyServer

AIProxyServer - Main Screen

AIProxyServer - Main Screen

Why AIProxyServer?

  • One vault, every provider — Store your ChatGPT, Claude, Gemini, Grok and other API keys once. Every connected app reuses them without ever seeing the secrets.
  • App Store friendly — Mobile and tablet apps distributed through stores cannot ship API key registration screens. Point those apps at AIProxyServer running on your Mac and the limitation disappears.
  • Drop-in OpenAI compatibility — Any client that speaks the OpenAI Chat Completions API works out of the box. No custom adapter required for Claude, Gemini, ERNIE, or any other supported cloud.
  • Encrypted on disk — API keys are stored in an encrypted vault using a per-machine master key with restricted file permissions.

Key Features

  • Eleven cloud providers — OpenAI (ChatGPT), Anthropic (Claude), Google (Gemini), xAI (Grok), Azure OpenAI (Copilot), Perplexity, Groq, DeepSeek, Moonshot (Kimi), Alibaba (Qwen), and Baidu (ERNIE)
  • OpenAI-compatible endpoints — /<provider>/v1/chat/completions and /<provider>/v1/models on every provider
  • Real-time streaming — Server-Sent Events (SSE) forwarded for every provider, including Anthropic and Gemini whose native formats are translated on the fly
  • Optional Bearer Token authentication — Off by default for hassle-free local use, easily switched on when you expose the proxy to your LAN
  • Localhost-only by default — Network exposure is opt-in; the proxy only binds to 127.0.0.1 until you enable LAN access
  • Encrypted credential vault — API keys and secret keys stored with Fernet encryption and a 0600 master key file
  • Auto-start on launch — Bring up the app and the proxy is immediately ready to receive requests
  • Per-provider notes and links — Quick access to each provider's API key page right from the main window

Perfect For

  • Developers building cross-platform apps that cannot embed API key registration UIs due to store policies
  • Flutter, Swift, Electron, and web app developers who want a single OpenAI-compatible backend for every cloud AI
  • Power users who subscribe to multiple AI services and want to switch providers without changing client code
  • Small teams sharing a single Mac as an AI gateway across phones, tablets, and laptops on the same Wi-Fi network
  • Anyone who would rather keep API keys in one vault on their own machine than scattered across every app's settings

Getting Started

  • Launch AIProxyServer — the proxy starts automatically on port 8421
  • Pick a provider from the list and click Set API Key, then paste the key obtained from the provider's console
  • For ERNIE add both API Key and Secret Key. For Azure OpenAI paste your deployment endpoint URL
  • Point your client app at http://localhost:8421/<provider>/v1 using the OpenAI SDK or any compatible HTTP client
  • If you want to access the proxy from another device on the same Wi-Fi, enable Allow LAN Access and turn on Require Bearer Token to protect it

Privacy

  • API keys never leave your computer. Stored encrypted with a master key in ~/Library/Application Support/AIProxyServer/ with 0600 permissions
  • The proxy only forwards requests to providers you explicitly configured — nothing is sent anywhere else
  • No telemetry, analytics, or background calls of any kind
  • No account, no sign-up, no cloud component. You run the proxy, you own the keys
  • Localhost-only listening by default; LAN exposure must be turned on by you