Claude Code model router
Gnosis
Terminal coding agent with browser UI, MCP, Obsidian memory, and live office floor. Any model via OpenRouter. MIT.
Install
npx @dominquechurch/gnosisFound in the project README. Check it for setup steps after installing.
git clone https://github.com/DOMCHURCH/Gnosis.gitFrom the README
Gnosis
Your AI agents needed an office. So I built one.
Gnosis is a terminal coding agent that runs on any model, switches models
mid-session without losing your history, and ships with a browser UI where you
watch your agents work as blocky figures moving around a 3D office floor.
Install
One command, in PowerShell:
It fetches the latest release, verifies the download, runs the installer, and
then tells you what else this machine needs for the optional parts.
Or take the installer directly:
Gnosis-Setup.exe (Windows, ~120MB)
Or just the CLI, if you have Node.js:
Passing options to the install script
iex cannot forward parameters to a piped script, so use the scriptblock form:
Desktop App
The desktop app includes:
- Three.js 3D office floor showing agents working in real time
- Integrated terminal panel
- Voice mode — say "hey jarvis" to activate (no API key required for wake word,
GROQ_API_KEY for transcription). "hey jarvis" is openWakeWord's closest
available model; a custom "hey gnosis" model is future work. The agent speaks
its reply back only for turns you started by voice — typing in the chat never
triggers speech
- Voice is built around the fact that the model round trip is the whole wait.
Transcription and speech are under a second each; everything else is one
request you are sitting through. So the route to the model is opened on the
wake word, while you are still talking, rather than inside your first
question — and a tool whose success speaks for itself ("open Spotify") is
confirmed the moment the tool returns instead of costing a second round trip
to be told what already happened. Per-turn timings are appended to
~/.dom/voice-timing.log if you want to see where yours go
- Voice is only as fast as the model you point it at,…Details
- Repository
- DOMCHURCH/Gnosis
- Language
- TypeScript
- License
- MIT
- Last updated
- 6 September 2026
- Topics
- ai-agentclaudecoding-agentdeepseekllmmcpopenrouterterminaltypescriptwindows
FAQ
What is Gnosis?
Terminal coding agent with browser UI, MCP, Obsidian memory, and live office floor. Any model via OpenRouter. MIT. It is a model router for Claude Code, from DOMCHURCH/Gnosis on GitHub.
How do I install Gnosis?
Run npx @dominquechurch/gnosis, then follow the setup in the README.
Is Gnosis popular?
DOMCHURCH/Gnosis has 31 GitHub stars and 5 forks. It was last updated on 6 September 2026.
More model routers & proxies
Hermes Claude Code Local
32KaiFelixBennett/hermes-claude-code-local
Run Hermes Agent + Claude Code locally on llama.cpp — zero API costs. A 4h / 7M-token session that would have cost $94 on Claude Opus 4.7
Claude Code Forge
29TensorBlock/claude-code-forge
Setup scripts for using TensorBlock Forge with Claude Code - access any AI model through Claude's interface
Ollama Herd
31geeks-accelerator/ollama-herd
Route LLMs, embeddings, image generation and speech-to-text across the Macs, Linux servers and Windows PCs you already own. One OpenAI-, Ollama- and Anthropic-compatible endpoint, so Claude Code and Codex run on your own machines. Zero config.
Zed API
29csmsqj/zed-api
将 Zed 托管模型转为本地 OpenAI / Anthropic 兼容 API(Responses、Ch Completions、Messages),支持流式与工具调用,Codex / Claude Code / OpenCode 可直连
Claude Code Proxy
32daodao97/claude-code-proxy
Claude code proxy, friendly view request log, enhance request stability
Claude2openai Proxy
30ziozzang/claude2openai-proxy
Claude API client to OpenAI API compatible API gateway/proxy. This can run Claude-Code with OpenAI API
Browse the most popular Claude Code mods, all model routers & proxies, or the skills and MCP server directories.