antondanv

Brainyard — Development skill for Claude Code

Development community

One API for agentic coding CLIs — Claude Code, Codex and Antigravity: status checks, one-shot asks, streamed agent runs with steering, sessions, models and MCP.

How to install Brainyard

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

What Brainyard does

One API for agentic coding CLIs — Claude Code, Codex and Antigravity: status checks, one-shot asks, streamed agent runs with steering, sessions, models and MCP.

Alternatives in Development

Claude API — Build apps with the Claude API or Anthropic SDK 94.1k ★
Gcli2api — 将 GeminiCLI 和 Antigravity 转换为 OpenAI 、GEMINI 和 Claude API 接口 4.6k ★
Antigravity Claude Proxy — Proxy that exposes Antigravity provided claude / gemini models, so we can use them in Claude Code and OpenClaw 3.5k ★

README

Brainyard

**One interface to the coding agents you already have: Claude Code, Codex and Antigravity.**

See which are installed and signed in, ask them one-shot questions, or run an agent in a folder and watch a live, human-readable feed of what it does. From the terminal, from TypeScript, or from a local dashboard.

[![CI](https://github.com/antondanv/brainyard/actions/workflows/ci.yml/badge.svg)](https://github.com/antondanv/brainyard/actions/workflows/ci.yml) [![npm](https://img.shields.io/npm/v/@antondanv/brainyard?color=6d7dfc)](https://www.npmjs.com/package/@antondanv/brainyard) ![node](https://img.shields.io/badge/node-%E2%89%A522-3c873a) ![dependencies](https://img.shields.io/badge/runtime%20dependencies-0-3c873a) [![license](https://img.shields.io/badge/license-MIT-blue)](LICENSE)

**English** · [Русский](README.ru.md)

The Brainyard dashboard: three status cards, and a Claude Code run that wrote fib.py, ran it and reported the output

Why

Claude Code, Codex and Antigravity are all excellent headless agents, and all three speak a different dialect: different flags, different stream formats, different ways to resume a session, pass an MCP server, pick an effort level or report a usage limit. Each also has quirks that only show up in real runs: a prompt starting with `---` read as a command-line option, a CLI that never exits because its stdin is still open, a turn that ends silently after a refused tool.

Brainyard gives them one API. The same options, the same events and the same errors for all three, with the workarounds built in. It drives the CLIs you installed with your own accounts. It runs no service, calls no model API and holds no keys.

It was extracted from a production system that runs all three CLIs every day. Most of the behaviour below exists because a real run failed without it ([battle-t