startup-idea-analyzer

CLI tool that scores startup ideas on a 0-60 scale and delivers a BUILD, PIVOT, or KILL verdict.

How it works

Three-stage AI pipeline using the Anthropic API:

Stage	Model	What it does
1. Structurer	Haiku	Decomposes your idea into problem, customer, solution, assumptions, business model. Generates search queries.
2. Researcher	Sonnet + web search	Finds real competitors, pricing, market size, and demand signals from the live web. Falls back to model knowledge if search fails.
3. Analyst	Opus	Scores 6 dimensions, identifies fatal flaws, and delivers the verdict.

Scoring dimensions (0-10 each)

• Problem Reality — Is this a real, painful problem people pay to solve?
• Niche Clarity — How well-defined and reachable is the target market?
• AI Advantage — Does AI provide a genuine, defensible edge?
• Solo Buildability — Can a small team ship an MVP in <3 months?
• Revenue Speed — How fast to first paying customer?
• Moat Strength — What prevents fast-followers from copying this?

Verdicts

• BUILD (40-60) — Strong fundamentals, go for it
• PIVOT (25-39) — Promising kernel, needs significant changes
• KILL (0-24) — Fatal flaws, move on

Setup

Requires Python 3.11+ and an ANTHROPIC_API_KEY environment variable.

uv sync
export ANTHROPIC_API_KEY=sk-ant-...

Usage

# Analyze an idea (with live web research)
startup-idea-analyzer "An AI tool that generates custom workout plans based on DNA analysis"

# Skip web search (faster, cheaper — uses model knowledge only)
startup-idea-analyzer -n "marketplace for fractional CTO services"

# Interactive mode
startup-idea-analyzer

Every run automatically exports three report formats to reports/<timestamp>_<slug>/:

• analysis.json — structured data
• analysis.md — markdown report
• analysis.html — standalone dark-themed HTML report

Live feedback

• Stage 2 streams research text in a live terminal panel with a token/cost counter
• Stage 3 shows a spinner with live token count and cost estimate
• Web search queries are displayed as they happen
• Ctrl+C between stages prints costs consumed so far

Cost

Typical analysis costs $0.10-0.30 depending on research depth. The CLI shows token usage and cost after each stage.

Stage	Model	Typical cost
Structurer	Haiku	~$0.003
Researcher	Sonnet	~$0.02-0.06
Analyst	Opus	~$0.08-0.25
Web searches	—	$0.01 each

Project structure

src/startup_idea_analyzer/
├── cli.py        # Typer CLI, streaming orchestration
├── pipeline.py   # 3-stage Anthropic API pipeline
├── prompts.py    # System prompts + tool schemas per stage
├── models.py     # Pydantic models, token/cost tracking
├── display.py    # Rich terminal output, StreamPanel
└── export.py     # JSON, Markdown, HTML report generation