Os3 Router — AI skill for Claude Code
Use your Codex (ChatGPT) or Claude Code subscription as the LLM for rabbit OS3: tool calling, workers, computer use, dashboard, watchdog.
How to install Os3 Router
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open Nesbesss/os3-router and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Os3 Router does
Use your Codex (ChatGPT) or Claude Code subscription as the LLM for rabbit OS3: tool calling, workers, computer use, dashboard, watchdog.
Alternatives in AI
- Vllm Mlx — High-performance OpenAI and Anthropic compatible LLM inference server for Apple Silicon 1.5k ★
- Room — Open-source earning-focused swarm intelligence engine 838 ★
- Image Use — Use your ChatGPT subscription to generate images from the command line — no OPENAI_API_KEY, no gateway, no dae 373 ★
README
![]()
os3-router
Use your **Codex / ChatGPT subscription** and/or your **Claude Code** login as the LLM for **rabbit OS3**: chat, tool calling, workers, browser and computer use, and **image generation and editing**, all on the subscription you already pay for. Mix them per role, e.g. `gpt-6-luna` for the main chat and `claude-sonnet-5` for the workers. It includes the OS3 Router app, token and limit tracking, a watchdog that repairs a stuck rabbit-agent, and automatic updates.
**Highlights**
- Your subscription, not an API bill: Codex (ChatGPT Plus/Pro) and Claude Code logins, per role
- Fast and light on your limits (0.5): one Codex stays running (replies up to 3× faster) and your chat keeps its conversation, so about 92% of it comes from the cache
- Two ChatGPT accounts (0.5): when one runs out, the router continues on the next and comes back later
- Images included: rabbit makes and edits pictures with Codex's own image generation, no paid image provider
- Browser and computer use that doesn't give up: workers look at the screen when a page can't be read (Excel Online, Google Docs, canvas apps) and check again before saying something isn't there
- Keeps working: watchdog for the rabbit-agent tunnel, fallback model at the usage limit, automatic updates that test themselves before switching over
OS3 cloud ──▶ rabbit-agent (your machine) ──▶ os3-router (localhost:11435) ──▶ codex exec / claude -p ──▶ your subscription
⚠️ **Read first.** This drives the official Codex CLI and/or the official Claude Code CLI, unmodified, with **your own** login. Check whether OpenAI's and Anthropic's terms allow using your subscription this way; that decision is yours (see [Claude Code](#claude-code) for what we found). Heavy agent use also burns through your 5-hour and weekly limits quickly (the dashboard shows both).
The project was called **codex-os3
Related Skills
Cheap LLM Router
Save Claude/Codex subscription tokens by routing context-only file reads through a cheaper LLM (DeepSeek, Kimi
JevXagent
When Jev meets LLM — a local LLM proxy & decision router for Claude Code (Anthropic API): telemetry, statistic
Prompt Router
Prompt-Router: lightning-fast LLM gateway with a telemetry dashboard for coding agents (Goose, Cursor, Claude
Agentic KVM
Give AI agents (Claude Code/Cowork, ChatGPT/Codex, Gemini/Antigravity) bare-metal, out-of-band control of any
Comfyui MCP Panel
The local-first sidebar AI agent for ComfyUI — runs on your own Claude OR ChatGPT subscription (no API keys, n
Vibe Bar
Native macOS menu-bar dashboard for AI subscription quotas, reset forecasts, local token costs, and provider s
Related Agents
GPT 5.6 Luna
General-purpose agent that runs on GPT-5.6-Luna (routed through the shunt gateway to the ChatGPT/Codex subscri
Codex Agent
Runs OpenAI Codex on the user's ChatGPT subscription. Three modes - consult (OS-sandboxed read-only analysis,
SaaS Builder
Builds multi-tenant SaaS products on Next.js, Supabase and Stripe with auth, roles, billing, an admin panel an