Back
2026

Meridian

Vague idea to clear plan with AI-assisted brainstorming

About

Meridian is a visual thinking workspace designed to turn vague starting points into structured branches worth pursuing. Starting from a single prompt, the app builds a branching idea tree through iterative exploration: each step generates grouped thematic questions, and committing one selected direction per question group creates a new tree node and advances the branch.

Three exploration personas — Architect, Writer, and Strategist — shape both the prompt structure and the metadata shown for each suggested direction, making the same idea feel different depending on whether you are pressure-testing architecture, refining narrative beats, or mapping execution risk.

The app operates in two modes. In manual mode, Meridian builds the exploration prompt, copies it to the clipboard, and waits for you to paste the LLM response back in. In in-app mode, a saved API key lets Meridian call OpenAI, Gemini, or Anthropic directly from the browser. A visual tree lets you navigate back to any earlier node to inspect or branch again, and the active path can be exported as a persona-shaped PRD-style plan: copy to clipboard, download as Markdown, or save as PDF. The full exploration tree, active node, persona selection, and provider settings persist in browser localStorage with no backend required.

Key Highlights
  • Builds a branching idea tree with grouped thematic questions per iteration
  • Three personas — Architect, Writer, and Strategist — shape both prompts and export structure
  • Two operation modes: manual clipboard flow and direct in-app calls to OpenAI, Gemini, or Anthropic
  • Visual tree navigation to inspect or re-branch from any earlier node
  • Exports the active branch as a persona-shaped PRD-style plan: copy, Markdown, or PDF
  • Full state persisted in browser localStorage with no backend required
Stack
VueTypeScriptViteTailwind CSS