Airelay

airelay

[中文](README_CN.md)

Use any LLM with Claude Code or Codex CLI — local protocol translation, zero setup.

**3.0 MB single binary. Starts in milliseconds. Zero runtime dependencies. macOS menu bar app.**

Supported Platforms

Platform	Status
macOS (Apple Silicon)	Full support — binary + tray
macOS (Intel)	Full support — binary + tray
Windows (x86_64)	Binary + CLI + server (tray not yet available)
Linux (x86_64)	Binary + CLI + server

What it does

Claude Code requires Anthropic's Claude models (paid subscription). Codex CLI requires OpenAI models.

airelay runs a local HTTP server that translates API protocols in real time — Anthropic Messages ↔ OpenAI Chat Completions, OpenAI Responses ↔ Chat — so you can use DeepSeek, Kimi, GLM, Qwen, Ollama, or any OpenAI-compatible API with your favorite AI coding tools.

Claude Code ── Anthropic Messages ──▶ airelay ── OpenAI Chat ──▶ DeepSeek / Kimi / ...
Codex CLI  ── OpenAI Responses ────▶                          ▶ Ollama / LM Studio / ...

Quick Start

1. Install

**macOS / Linux**

curl -fsSL https://raw.githubusercontent.com/ligj1706/airelay/main/install.sh | bash

**Windows (PowerShell)**

irm https://raw.githubusercontent.com/ligj1706/airelay/main/install.ps1 | iex

After installation, **close and reopen your terminal** (or run `exec $SHELL` on macOS/Linux) to activate the commands.

The install script handles three things: puts `airelay` in `~/.local/bin/`, appends to PATH, and writes the `ar` alias.

If you prefer building from source (requires Rust):

git clone https://github.com/ligj1706/airelay.git
cd airelay && cargo build --release
cp target/release/airelay ~/.local/bin/

2. Start

ar

The install script sets up the `ar` alias: it checks whether airelay is running and starts it in the background if not. Then open `http://127.0.0.1:8082/admin` to configure.

3.