Net Promoter Score (NPS) Intelligence — MCP for Claude

A remote MCP server that lets Claude (and any other MCP-aware client) query a curated database of publicly disclosed Net Promoter Score (NPS) values from named companies — with source URLs, news snippets, and pre-computed sector benchmarks across 1,000+ industries.

3,806

NPS Disclosures

1,097

Sector Benchmarks

15 yrs

2011 – Today

~few days

Refresh Cadence

Try asking Claude

What has Salesforce publicly disclosed about NPS?

Show me Banking NPS disclosures since 2023, highest first.

What's the median publicly disclosed NPS in Insurance? How many companies?

Which SaaS companies have the highest publicly disclosed NPS, and where did they publish it?

Has Tesla disclosed an NPS? Show me the trend by year.

What does this database cover? How far back does it go?

Tools exposed

lookup_company — every disclosed NPS for a named company, with citations
search_disclosures — filter by sector, year range, and quality
get_sector_benchmark — pre-computed mean / median / p75 / min / max per sector
list_coverage — what the database actually contains

Connect

Claude.ai (browser)

Settings → Connectors → Add custom connector.

https://cg-nps-mcp.adam-dorrell.workers.dev/mcp

Claude Desktop (Mac/Win)

Uses mcp-remote proxy with the /sse endpoint.

https://cg-nps-mcp.adam-dorrell.workers.dev/sse

Power users — install the skill

The MCP server ships its own instructions automatically — every client that connects gets the citation rules, tool guidance, and disclosure-bias framing without any extra install. But if you want the same rules applied locally (e.g. when working with an exported CSV), grab the skill as a markdown file:

curl -o ~/.claude/skills/cg-nps/SKILL.md https://cg-nps-mcp.adam-dorrell.workers.dev/skill

Why this is interesting

Claude doesn't know what specific companies have published about their NPS. This MCP gives it 15 years of curated public disclosures with a source URL on every row — so any answer Claude gives can be verified, and any company named is grounded in a real news article or investor document. It's the evidence layer that turns "what's a typical NPS?" from a guess into a citation.