xiongdi

Cab

Name: Cab
Author: xiongdi

AI community

SOURCE https://github.com/xiongdi/cab

Description

One local gateway for every coding agent. Cost-aware LLM model routing for Claude Code, Codex, OpenCode, and more — rank by intelligence, coding ability, price, or speed across OpenAI/Anthropic providers.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

CAB — Coding Agents Bridge

**One local gateway for every coding agent. Smart, cost-aware LLM routing.**

[English](README.md) · [简体中文](README.zh-CN.md) · [Documentation](https://xiongdi.github.io/cab/) · [Changelog](https://xiongdi.github.io/cab/project/changelog/)

[![Release](https://img.shields.io/github/v/release/xiongdi/cab?color=brightgreen&label=release)](https://github.com/xiongdi/cab/releases) [![Downloads](https://img.shields.io/github/downloads/xiongdi/cab/total?color=blue&label=downloads)](https://github.com/xiongdi/cab/releases) [![Platform](https://img.shields.io/badge/platform-Windows%20%7C%20macOS%20%7C%20Linux-lightgrey)](https://github.com/xiongdi/cab/releases) [![License](https://img.shields.io/github/license/xiongdi/cab?color=orange)](LICENSE) [![CI](https://img.shields.io/github/actions/workflow/status/xiongdi/cab/build-cli.yml?label=build)](https://github.com/xiongdi/cab/actions/workflows/build-cli.yml)

CAB sits between your coding agent CLIs and upstream LLM providers — a single local gateway at `http://localhost:3125/v1`. Point Claude Code, Codex, OpenCode, or Grok Build at it; CAB ranks every request to the best enabled model for the job and forwards it to OpenAI / Anthropic. No more hand-editing JSON, TOML, or `.env` files, no more guessing which model is cheapest or fastest.

**Why CAB?** Coding agents each expect their own API keys, endpoints, and model lists. CAB gives you **one endpoint, one key, one dashboard** for all of them — and routes each prompt by capability, benchmarks, and real token cost.