Live instance: businesslog-ai
You have a raw business idea. You want structured analysis, not cheerleading. This agent applies Porter's Five Forces and unit economics to your idea. It runs on your infrastructure. You control the data.
Most AI analysis tools are either locked behind subscriptions or send your data to third parties. This agent runs standard business frameworks without requiring a login or sharing your notes. The analysis is yours.
- Fork this repository. You own the code.
- Deploy it to your Cloudflare account. It runs as a Cloudflare Worker.
- Connect your own LLM API key. The worker sends prompts directly to your chosen endpoint (e.g., OpenAI, Anthropic, or a local model via Cloudflare AI).
- You get a structured report. The agent formats the LLM's output into the requested business framework.
There is no intermediary server. Your API key is stored as a Cloudflare Secret. Conversation context is kept only in your browser session.
- Fork this repository to your own GitHub account.
- Install and authenticate Wrangler, Cloudflare's CLI tool.
- Deploy the Worker:
npx wrangler deploy
- In the Cloudflare dashboard, set your LLM API key as a secret named
AI_API_KEY. The worker includes a simple adapter for common OpenAI-compatible endpoints.
- Applies Standard Frameworks: Generates analysis using Porter's Five Forces and basic unit economics for a given idea.
- Self-Hosted: The code runs on your Cloudflare Worker. Your prompts and API key never pass through our servers.
- LLM Agnostic: Configure it to use any endpoint with an OpenAI-compatible API.
- Zero Dependencies: The Worker script has no npm dependencies.
- Session-Based Context: Maintains conversation history only within your current browser tab.
- Fleet-Compatible: Can receive updates if you connect it to the Cocapn Fleet.
- Analysis Depth is LLM-Dependent. The quality and accuracy of the generated report are bound by the capability of the language model you connect it to. It will not perform independent research or fact-checking.
- Response time is variable. The ~3 second claim depends entirely on your configured LLM's speed. Network latency and model processing time will affect this.
MIT License. You are free to use, modify, and distribute this software.
Built by Superinstance and Lucineer (DiGennaro et al.).