dorukardahan

Headless Relay — AI skill for Claude Code

AI community

An Agent Skill that lets your coding agent ask the other AI models on your machine for help: GPT, GLM, Grok, Gemini, Claude, even local models.

How to install Headless Relay

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

What Headless Relay does

An Agent Skill that lets your coding agent ask the other AI models on your machine for help: GPT, GLM, Grok, Gemini, Claude, even local models. Second opinions, parallel consensus, image generation. Works in Claude Code, Codex, Grok, Cursor, OpenClaw, Hermes and friends.

Alternatives in AI

Define — Definition phase - Clarify and scope problems with multi-AI consensus 2.8k ★
HolyClaude — AI coding workstation: Claude Code + web UI + 7 AI CLIs + headless browser + 50+ tools 2k ★
Rb Setup — First-time setup 1.3k ★

README

headless-relay

headless-relay is an [Agent Skill](https://agentskills.io) that lets your coding agent use the other AI models installed on your machine, without you leaving the session. You say "ask Codex what it thinks of this function", "get a second opinion on this bug from GLM and Grok", or "have Gemini generate an image for this post". Your agent quietly runs the right tool in the background, reads the answer, and reports back to you. No new accounts, no API juggling: it drives the AI tools you already have, with the logins you already use.

It works in any coding agent that reads Agent Skills and can run shell commands: Claude Code, OpenAI Codex CLI, xAI Grok, Cursor, OpenClaw, Nous Research Hermes and friends. Five model lanes ship ready to use (GPT, GLM, Grok, Gemini, Claude), and you can plug in your own, including local models running through Ollama, LM Studio, or Apple MLX. Installation is a single `git clone`; everything else on this page is detail for when you need it.

⚠️ Security: the Grok lane and your repository

Earlier shipped versions of the xAI Grok Build CLI, run inside a git repository, uploaded your **entire repo: full git history and all tracked files, including a tracked `.env`** to xAI cloud storage, regardless of which files the model reads — confirmed by xAI's own Grok account on X and by independent wire capture. On 2026-07-15 xAI open-sourced Grok Build (Apache-2.0), deleted previously-retained coding data, and set retention off by default. A source audit of the released code (commit `c68e39f`) found the whole-repo bundle path is **gone** from source, the remaining trace-upload pipeline (the model's own turn I/O) defaults **off**, and your local `~/.grok/config.toml` beats xAI's remote settings — xAI cannot remotely re-enable an upload you've turned off locally. Two residuals remain, and this skill still guards both: Grok auto-loads global rules into every turn — your other tools' config (a Claude-Code/C