enx

enx is an OpenAI-compatible LLM proxy gateway. It exposes a single endpoint that speaks the OpenAI and Anthropic wire formats, normalizes every request into one internal representation, and forwards it to the upstream provider you route it to. It ships as a single binary that serves the API, the management UI, and the account pool on one port.

• Website: https://enowxlabs.com
• Community: https://discord.gg/enowxlabs

Installation

Install script

Linux and macOS:

curl -fsSL https://raw.githubusercontent.com/enowdev/enowx/main/install.sh | sh

Windows (PowerShell):

irm https://raw.githubusercontent.com/enowdev/enowx/main/install.ps1 | iex

The installer downloads the latest release for your OS/arch, verifies its checksum, and installs the enx binary (to /usr/local/bin on Unix, or %LOCALAPPDATA%\Programs\enx on Windows). Override the location with ENX_INSTALL_DIR, or pin a version with ENX_VERSION=vX.Y.Z.

Download a release binary

Prebuilt binaries for Linux, macOS, and Windows (amd64 and arm64) are attached to every release on the Releases page. Download the asset for your platform, rename it to enx (or enx.exe on Windows), make it executable, and place it on your PATH.

Build from source

Requires Go 1.26+ and Node 22+.

git clone https://github.com/enowdev/enowx.git
cd enowx
make build      # builds the web UI, embeds it, and produces bin/enx
./bin/enx

Usage

Start the gateway:

enx

By default it listens on 127.0.0.1:1430. Open http://localhost:1430 for the management UI, where you add provider accounts (keys/tokens) to the pool.

Once an account is added, send a standard OpenAI request:

curl http://localhost:1430/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Streaming is supported with "stream": true and is returned as OpenAI-style server-sent events.

Routing

The model field selects the upstream provider:

• A provider/model prefix routes explicitly, for example codebuddy/... or kiro/....
• Known prefixes route automatically (kiro-..., codebuddy-...).
• Anything else falls back to the OpenAI-compatible upstream.

Configuration

Configuration is read from the environment and an optional config.json in the runtime directory.

Variable	Default	Description
ENOWX_PORT	1430	Listen port.
ENOWX_HOST	127.0.0.1	Listen address.
ENOWX_RUNTIME_DIR	~/.enowx	Data directory (SQLite database).
ENOWX_LOG_LEVEL	info	Log verbosity.

State is stored locally in a pure-Go SQLite database; no external services are required to run the gateway.

Commands

enx            # start the gateway
enx version    # print the version

Endpoints

• POST /v1/chat/completions — OpenAI-compatible chat completions.
• GET /health — health check.
• GET /api/* — management API used by the UI.
• / — embedded management UI.

Providers

enx normalizes inbound OpenAI and Anthropic traffic into a single internal request. Outbound, each provider re-encodes only what it needs: providers that already speak OpenAI pass through unchanged, while providers with their own formats are normalized per provider. Current providers:

• OpenAI-compatible upstreams
• CodeBuddy
• Kiro

Development

Linux and macOS:

./dev.sh

Windows (PowerShell):

./dev.ps1

This installs dependencies if needed, then runs the backend and frontend on one port (http://localhost:1430) with hot reload and no build step. The Go server proxies the SPA and its hot-reload channel to an internal Vite dev server.

The built-in terminal uses a PTY and currently works on Linux and macOS; on Windows the rest of the app runs but the terminal is disabled.

Community and support

• Website: https://enowxlabs.com
• Discord: https://discord.gg/enowxlabs

License

See LICENSE.