vtemian/octto

What it solves

Octto replaces the tedious process of typing long, detailed prompts or answering sequential terminal-based questions when brainstorming with AI. It transforms the AI's request for information into an interactive, visual browser UI, reducing the time spent typing from minutes to seconds.

How it works

The system uses three specialized agents—a bootstrapper, a probe, and an orchestrator—to manage the session. The bootstrapper splits a user's request into multiple exploration branches. The probe agent determines if a branch requires further questions to refine the requirement. These questions are then presented to the user via a browser interface featuring 14 different input types (such as radio buttons, sliders, and code editors).

Who it’s for

Developers and designers who use AI agents for planning and brainstorming and want a more efficient, visual way to interact with their project requirements.

Highlights

• Visual Input Types: Supports 14 rich input types including rank reordering, star ratings, and side-by-side code diffs.
• Parallel Exploration: Splits requests into 2-4 branches that can be answered in any order.
• Live Updates: Questions appear instantly as previous ones are answered, with a progress indicator for the remaining steps.
• Customizable Agent Behavior: Allows injection of custom instructions (fragments) via global or project-specific configuration files.
• Final Plan Generation: Renders a final plan for review and that is saved directly to the project's documentation folder.