# Demand Discovery AI

> Demand Discovery AI is a startup-idea validation platform built on the principle "behavior over opinion." Instead of measuring stated interest (surveys, friend feedback, LLM guesses), it measures revealed behavior — search intent, outreach replies, landing-page engagement, and buying signals — and produces a 0-100 Demand Score with a clear Build / Pivot / Kill verdict. This site hosts the public MCP (Model Context Protocol) server that exposes the Demand Discovery sales-agent to any MCP-compatible AI client (Claude Code, Claude Desktop, Cursor, Devin Desktop, ChatGPT). The MCP runs at https://mcp.demanddiscovery.ai/api/mcp via Streamable HTTP, with no auth required.

## Product

- [Demand Discovery AI main site](https://demanddiscovery.ai): the product itself. Starter ($49 — vets two ideas end to end), Founder Pack (5 idea slots), Studio Pack (25 idea slots). Slot-based portfolio model.
- [Free Market Research entry](https://demanddiscovery.ai/free-report): Step 1 of every report bundle, free of charge.

## MCP server

- [MCP endpoint](https://mcp.demanddiscovery.ai/api/mcp): canonical Streamable HTTP MCP URL for installation in any compatible client.
- [MCP manifest](https://mcp.demanddiscovery.ai/api/mcp/manifest): JSON manifest with tool list, version, and metadata.
- [Agent install page](https://mcp.demanddiscovery.ai/agent/): plain-HTML step-by-step install instructions written for AI agents — fetch it and follow the section for your client. Includes a curl verification call.
- [Install snippets](https://mcp.demanddiscovery.ai): copy-paste install commands for Claude Code, Claude Desktop, Cursor, Devin Desktop, and ChatGPT.
- [Claude Skill](https://mcp.demanddiscovery.ai/#skill): optional free Skill for Claude clients that guides the full validation flow end to end. Claude Code: /plugin marketplace add demand-discovery-ai/mcp. Claude Desktop: upload the skill zip from https://mcp.demanddiscovery.ai/demand-discovery-skill.zip under Settings -> Capabilities -> Skills.
- [Skills CLI](https://github.com/demand-discovery-ai/mcp): npx skills add demand-discovery-ai/mcp — one command installs the Demand Discovery Skill into 70+ coding agents (Cursor, Codex, OpenCode, Claude Code, and more).
- [Vibe-coding agents](https://mcp.demanddiscovery.ai/#agents): connect Demand Discovery to Replit Agent (skill file), Lovable (custom chat connector), v0 by Vercel (custom MCP connection), Bolt (custom MCP connector), Base44 (custom MCP connection), or a whole Cursor repo (.cursor/mcp.json) using the same public MCP URL.

## Framework

The Demand Discovery framework is a three-step bundle:

1. Market Research Report — maps the landscape (TAM/SAM/SOM, competitive landscape, GTM, growth drivers).
2. Demand Discovery Report — tests behavior with hosted landing-page tests, AI-defined ICPs, AI-written outreach to real prospects, reply analysis, and the 0-100 Demand Score plus Build/Pivot/Kill verdict.
3. Agentic Launch — 90 days of continuous outreach to real ICPs, refining ICP and messaging from real conversations.

## Data sources

The Demand Score triangulates four behavioral buckets across 40+ data sources:

1. Search & attention — what people are searching for, watching, and clicking right now.
2. Conversation & pain — where people articulate problems, vent about workarounds, and ask each other for help.
3. Adoption & spend — the hardest signal to fake: installs, downloads, paid usage, reviews, marketplace traction.
4. Capital & hiring — funding momentum, hiring velocity, competitive tech adoption.

Passive scoring tools triangulate buckets 1 and 2 only. Demand Discovery triangulates all four, then closes the loop with active human signal via Agentic Launch. The specific connector list isn't public for competitive reasons — what matters is the four-bucket structure.

## Brand voice

"Behavior over opinion." Most founders are validating interest, not demand. The right question is not "would you use this" but "who is acting like they want this right now?" Real signals from real people. Not LLM guesses.