LLM Coding Bridge — AI skill for Claude Code

@sevoniva/llm-coding-bridge

[](https://www.npmjs.com/package/@sevoniva/llm-coding-bridge) [](https://github.com/sevoniva/llm-coding-bridge/actions/workflows/ci.yml) [](LICENSE)

A local, production-oriented protocol bridge for coding clients that need one stable OpenAI-compatible endpoint in front of one or more upstream models.

The bridge keeps client configuration, model aliases, upstream model IDs, credentials, streaming behavior, and route health in one controlled local service. It does not modify client source code and it does not store upstream API keys in the bridge configuration.

At a glance

Codex / Claude / OpenAI-compatible client
                |
                v
      http://127.0.0.1:37629/v1
                |
      LLM Coding Bridge
      - route by model alias
      - normalize protocols
      - enforce timeouts and limits
      - emit streaming heartbeats
                |
                v
        OpenAI-compatible upstream

Supported client surfaces:

Client or protocol	Bridge endpoint	Primary use
Codex CLI / Codex Desktop	/v1/responses	Responses API
Claude-compatible clients	/v1/messages	Anthropic-compatible messages
OpenAI-compatible clients	/v1/chat/completions	Chat Completions
Client discovery	/v1/models, /health	Model and service checks

The default listener is loopback-only. The service has no runtime dependencies and can be managed by macOS launchd.

Install

Requires Node.js 18 or newer.

npm install -g @sevoniva/llm-coding-bridge@latest
llm-coding-bridge --help

Quick start

Run the guided setup. It creates `~/.llm-coding-bridge/config.json`, keeps credentials outside that file, and can optionally configure local clients.