KDP Niche Scorer MCP

AI-native Amazon KDP niche scoring — feed a keyword, get BSR, competitor count, page count, review velocity, and estimated monthly sales as one opinionated score.

Built for self-published Kindle authors who want their AI agent (Claude, ChatGPT, Cursor) to pick their next book niche on real data — not vibes. Competes with Publisher Rocket ($199), KDSPY ($47), Helium 10 ($39–249/mo). This one lives inside your LLM workflow.

Install (hosted on MCPize)

The fastest way to use it — no local setup, OAuth handled for you:

# Claude Code
claude mcp add --transport http kdp-niche-scorer https://kdp-niche-scorer-mcp.mcpize.run/mcp

# Cursor
cursor mcp add kdp-niche-scorer https://kdp-niche-scorer-mcp.mcpize.run/mcp

# Windsurf
windsurf mcp add kdp-niche-scorer https://kdp-niche-scorer-mcp.mcpize.run/mcp

Or grab it from the marketplace: https://mcpize.com/mcp/kdp-niche-scorer-mcp

Pricing (hosted)

• Free — 5 requests/day. Enough to score a couple of niches per session.
• Pro $9/mo — 30 requests/day (~900/mo). For active indie authors.
• Author $29/mo — 150 requests/day (~4,500/mo). Multi-niche / agency use.

Tools

Scoring formula

Ported verbatim from the KDP Research Chrome extension. Out of 100:

• Demand (35) — share of top 10 selling well by BSR. Falls back to review-count signals when BSR isn't exposed by the search endpoint.
• Competition (30) — median review count of top 10 (fewer = easier).
• Pricing (15) — average price in the $5.99–$9.99 low-content sweet spot.
• Indie + Recency (10) — indie wins + recent publication bonuses.
• Top-3 Dominance (−25) — if top 3 books hoard sales vs positions 4–10.

Bands: Excellent (85+), Great (70+), Good (55+), Okay (40+), Weak (25+), Poor (<25).

Requirements

• Python 3.11+
• uv
• A RapidAPI key for the Real-Time Amazon Data API. Free tier = 100 req/mo, no card.

Quick Start

# Install dependencies
uv sync

# Set up your .env
cp .env.example .env
# edit .env and paste your RAPIDAPI_KEY

# Start the server with hot reload + interactive playground
mcpize dev --playground

Configuration

Development

uv run pytest            # Unit tests (22 offline tests covering the scoring formula)
uv run ruff check .      # Lint
uv run ruff format .     # Format
mcpize dev               # Hot-reload server on :3000
mcpize dev --playground  # Same + browser playground via public tunnel
mcpize doctor            # Sanity-check the build before deploy
bash test-mcp.sh         # MCP protocol smoke test (server must be running)

Deploy

mcpize login                                # once
mcpize secrets set RAPIDAPI_KEY your-key    # set production secret
mcpize deploy                               # ship

Data sources

v1 (this server): RapidAPI Real-Time Amazon Data /search endpoint only. Gets title, price, review count, rating per result. BSR is not returned by the search endpoint — the scoring formula's built-in review-based demand fallback handles this gracefully.

v2 (future): Keepa API for historical BSR + review velocity; Amazon Creators API once eligible.

Example

score_niche(keyword="low content journal", marketplace="US")
→ {
    score: 62,
    band: "Good",
    verdict: "worth_considering",
    signals: { demand: 21, competition: 25, pricing: 15, indie_recency: 1, penalty: 0, ... },
    top_competitors: [ { asin, title, price, review_count, est_monthly_revenue }, ... ]
  }

License

MIT