Documentation

Reasoning Architect turns a rough idea into a structured, tested prompt. These pages explain the workflow and the reasoning frameworks it can build on.

How the workflow runs

ResearchEnter a topic, upload files or connect a GitHub repo. The research step analyzes it and recommends a reasoning framework, with the reason.
BuildFill in a structured prompt field by field: role, task, context, constraints and output format, with suggestions at each step.
VariablesAccept or edit suggested {{placeholders}} so the finished prompt works as a reusable template.
RefineAI rewrites the weak parts, explains each change, and scores the prompt before and after on five criteria.
Test runRun the prompt on any Gemini model your API key can access, with temperature, top-p and length controls.

Guides

Reasoning frameworks

How each of the 6 frameworks reasons, the exact protocol it adds to your prompt, and what it costs.

Read the guide

Choosing a framework

A short decision guide and a side-by-side table of speed, cost and prerequisites.

Compare frameworks

What you need

Account: Sign in with Google. Access is limited to approved accounts.
Gemini API key: Add your own key in Settings. Model choice follows what the key can access.
Files (optional): PDF, Excel (.xlsx, .xls), CSV, TXT, Markdown and email (.eml).
GitHub (optional): Connect a repository and the most relevant files are packed into context.
Storage: Prompt history and usage stats stay in your browser. History exports as JSON.

Start building