Decomposer

Overview

Decomposer is an AI planning tool that takes a sprawling, multi-part goal and turns it into 8-14 concrete task cards -- each with a title, specific action items, dependencies, and a paste-ready sub-prompt you can run in any AI tool.

Every card gets an honest effort label: You (👤), AI (🤖), or Expert (🧰) -- so you know exactly what to delegate, what to do yourself, and when to call in a specialist. A built-in evaluator scores the plan across five quality dimensions and flags gaps before you start working.

Plans are persistent, shareable, and editable. As you complete tasks, outputs flow into dependent cards so downstream steps refine themselves automatically.

Why we built it

People bring big goals to AI and get paragraphs that sound smart but collapse during execution. Steps get missed, tasks overlap, and nothing tells you which parts actually need a human. You shouldn't need prompt-engineering skills to get a plan you can follow.

Decomposer sits upstream of execution. It makes the problem legible, routes work honestly, and hands you structured prompts -- so when you do open your preferred AI tool, you already know exactly what to ask.

How it works

1. Describe your goal -- Paste your objective and add context. Upload documents if you have them (PDF, Word, images, Markdown). Decomposer forms an observed brief and surfaces hidden assumptions you can accept or edit. 2. Get a structured plan -- The AI generates 8-14 task cards with titles, action items, effort labels, dependencies, and copy-ready sub-prompts. 3. Review quality -- A built-in evaluator scores coverage, redundancy, actionability, label routing, and synthesis fit. Low scores trigger targeted regeneration. 4. Execute and refine -- Track status per card, edit inline, capture outputs. Completed task outputs propagate to dependent cards so downstream prompts sharpen automatically. 5. Take it anywhere -- Copy sub-prompts into your preferred AI tool. Export to PDF or Markdown. Share persistent links with teammates.

Feature highlights

• Planning -- Works across any domain with no rigid templates
• Planning -- Surfaces hidden assumptions and constraints before you start
• Planning -- Coherent plan summary with cross-cutting risk identification
• Planning -- Granularity coaching flags tasks that are too broad or too narrow
• Effort Routing & Quality -- Honest labels: You (👤) / AI (🤖) / Expert (🧰)
• Effort Routing & Quality -- Five-dimension quality evaluation with targeted regeneration
• Effort Routing & Quality -- Guided execution modes (autonomous / guided / manual) with domain-aware artifacts
• Execution -- Dependency flow: completed outputs auto-refine downstream tasks
• Execution -- Inline editing, status tracking, and plan history with auto-save
• Execution -- Document upload: PDF, Word, images, text, Markdown
• Portability -- Copy-ready sub-prompts for ChatGPT, Claude, Gemini, or local models
• Portability -- PDF export with professional formatting
• Portability -- Markdown export for any tool
• Portability -- Persistent shareable links
• Portability -- Private by default -- we don't train on your data

Pricing

Free -- 3 plans. Pro -- $10/month -- Unlimited plans, PDF export, priority generation.

Notes

• Status: Production (v1.0) -- authentication, billing, document upload, mobile-first, dependency tracking
• Model-Agnostic: Thin adapter layer allows model swapping as quality and cost evolve
• No Lock-In: Everything exports cleanly