Cubiczan Server

Connect your Off Grid mobile app to Cubiczan's multi-agent intelligence.

What it does

Exposes Cubiczan's agent orchestration capabilities (CHP consensus, multi-agent workflows, stigmergic coordination) as an OpenAI-compatible API endpoint that Off Grid phones discover via LAN or connect to manually.

Quick Start

pip install -r requirements.txt
export CUBICZAN_API_KEY="your-key"
python server.py

The server starts on 0.0.0.0:8080 and implements the OpenAI /v1/chat/completions endpoint.

How it works

Off Grid Client	Cubiczan MCP Server
Remote LLM Server (OpenAI-compatible)	FastAPI serving /v1/models + /v1/chat/completions
Tool calling enabled	Returns function calls for Cubiczan agents
LAN discovery (port 8080)	Responds to HTTP probes on /v1/models
Secure keychain API key storage	Validates API key from Authorization header

Available Tools

The server exposes Cubiczan's agent capabilities as OpenAI function tools:

• cubiczan_analyze — Run multi-agent analysis on a problem
• cubiczan_consensus — Run CHP governance consensus on a proposal
• cubiczan_swarm — Deploy a stigmergic swarm for continuous monitoring
• cubiczan_search — Search Cubiczan knowledge base
• web_search — Real web search
• read_url — Fetch and summarize a URL

Adding to Off Grid

1. Open Off Grid → Settings → Remote Servers → Add Server
2. Enter http://<host>:8080 as the endpoint
3. Enter your Cubiczan API key
4. The model cubiczan-agent will appear in your model list
5. Enable tool calling in chat settings