🖥 Self-hosted · Port 3001 · GB10 fleet

One tab for your
entire AI stack

Self-hosted operator console for local AI infrastructure. Unified chat, web terminal, VS Code, AIKB memory search, and Ollama model management — all in a single browser tab.

Deploy GitHub →
Surfaces
What's inside

Each surface is a tab in the browser. No separate apps, no context switching.

💬
Chat
Unified chat interface across Claude, Gemini, and Codex. Switch models without leaving the tab.
Live
⬛
Terminal
Full web terminal via ttyd. SSH directly into your GB10 units from the browser.
Live
📝
VS Code
code-server running on the target machine. Full editor, extensions, git — browser-native.
Live
🤖
Model Manager
Pull, inspect, and remove Ollama models across your GB10 fleet. GPU metrics and HuggingFace search built in.
GB10
🧠
Memory
AIKB search and proposal review lane — surface what your AI agents have learned across sessions.
Live
Stack
What's under the hood
ComponentRole
Node.js / ExpressBackend proxy and API layer
ttydWeb terminal — exposes a shell over WebSocket
code-serverVS Code in the browser, runs on the target host
Ollama APIModel pull, list, delete — local inference on GB10
AIKB runtime eventsMemory search and proposal promotion
NginxReverse proxy — all surfaces under one domain/port
Deploy
Get AI Hub running

Designed to run on a dedicated VM or any always-on Linux host on your local network.

git clone https://github.com/mcglothi/ai-hub.git cd ai-hub && docker compose up -d
Web UI at http://<host>:3001 after startup
git clone https://github.com/mcglothi/ai-hub.git cd ai-hub && npm install cp .env.example .env # edit API keys npm start
Requires Node.js 18+, ttyd, and code-server on PATH