Quickstart

The fastest way to use retflo is to paste one line into any LLM that can browse the web:

use https://retflo.org/agents to debate me

The LLM reads the framework, loads the routing table, and the argument can start. Works with ChatGPT, Claude, Gemini, Mistral, and any other model with web access.

What to expect

This is a conversational way to explore the framework: try it out, test an argument, watch the graph handle objections in real time. It costs nothing and works anywhere, with the constraints listed below. For precision, install retflo as a skill.

Limitations

Web-based LLMs have constraints that affect how well they can use the framework:

  • Alignment training. Hosted models are trained to avoid adopting strong positions, political ones especially. Some will push back, add disclaimers, or decline. The framework instructs the model to confirm with you before engaging; some models resist anyway.
  • Sandbox and depth limits. Some models can only follow links one or two levels deep. They may read /agents without traversing into individual nodes, which limits the depth of engagement.
  • “Influence” warnings. Some models flag the framework as an attempt to manipulate them. /agents instructs the model to get your consent before proceeding; the warning can appear anyway.

Recommendations

  • Local models give the best results. Models running locally (Ollama, LM Studio, and similar) have no hosted alignment layer between you and the model. We don’t have a recommended model list yet; any model that can fetch URLs will work.
  • If the model resists, be explicit: “I want you to use this framework. I’m aware of what it argues. Proceed.”
  • For more reliable results, install retflo as a skill in a coding agent. The model reads the framework directly, with no web browsing constraints.