CCX

CCX is an open-source AI API proxy and protocol translation gateway built in Go by BenedictKing. It provides a single unified entrypoint for routing requests to Claude, OpenAI Chat, OpenAI Images, Codex Responses, and Gemini APIs, with an embedded web admin console and single-binary deployment. The project is licensed under MIT and available on GitHub, with the latest release at v2.8.19.

What It Is

CCX sits between AI clients and upstream API providers, acting as a reverse proxy and protocol translator. It normalizes requests across different AI API formats — Claude Messages, OpenAI Chat Completions, OpenAI Images, Codex Responses, and Gemini — exposing them all through a single port. This makes it useful for developers who want to consolidate multiple AI provider keys, implement failover, or route traffic across providers without changing client code.

Architecture and Deployment Model

CCX uses an integrated backend + frontend architecture where the frontend is embedded directly into the binary, enabling single-port, single-binary deployment. The backend exposes one entrypoint on port 3000 (configurable), routing traffic to the web UI, admin API, and all AI proxy endpoints. Deployment options include:

• Binary: Download a prebuilt binary and configure via .env file
• Docker: Official Docker image with Docker Compose support and optional Watchtower auto-update
• CCX Desktop: A desktop application available on Windows (Microsoft Store) and macOS (Homebrew cask), with code signing via SignPath Foundation
• Build from source: Requires Go 1.25+, Bun, and Make

Channel Orchestration and Smart Scheduling

A core feature of CCX is its channel management system. Operators can configure multiple upstream channels (API providers or keys), each with:

• Per-channel API key rotation
• Priority scheduling and promotion windows
• Health checks, circuit breaker, and automatic failover
• Custom headers, model allowlists, and route prefixes
• Proxy support per channel

The web admin console provides drag-and-drop priority adjustment, real-time traffic stats (request count, success rate, latency), and log viewing.

Supported Protocols and Endpoints

CCX proxies the following API protocols through a unified entrypoint:

• POST /v1/messages — Claude Messages
• POST /v1/chat/completions — OpenAI Chat Completions
• POST /v1/responses — Codex Responses (with session tracking for multi-turn workflows)
• POST /v1/images/generations, /edits, /variations — OpenAI Images
• POST /v1beta/models/{model}:generateContent — Gemini
• GET /v1/models — Models API

The project documentation also lists support for additional upstream providers including DeepSeek, Kimi, GLM, MiniMax, and Tongyi Qianwen via the channel configuration system.

Update: v2.8.19

The latest release is v2.8.19, published on 2026-05-30. The repository was created in January 2026 and has seen rapid development, accumulating over 2,700 GitHub stars and 200 forks within months. Active development continues with regular releases, and the project maintains an open issues tracker and community QQ group for discussion.