I Built an AI That Tells You If Your Idea Already Exists — And Syncs Results to Notion

#devchallenge #notionchallenge #mcp #ai

Notion MCP Challenge Submission 🧠

This is a submission for the Notion MCP Challenge

What I Built

Idea Reality Tracker — a dual-MCP pipeline that validates software ideas against 5 live platforms and automatically syncs structured results to a Notion database.

Instead of googling "has anyone built this?" and drowning in 10 tabs of noise, you describe your idea in one sentence. In 15 seconds you get:

A Reality Score (0–100) measuring how crowded the space is
Market momentum analysis (accelerating / stable / declining)
Competitor counts from GitHub, npm, PyPI, Hacker News, and Product Hunt
A Build / Pivot / Kill recommendation

All automatically saved to your Notion workspace as a searchable decision log.

The Story Behind It

Six months ago, I asked ChatGPT if my idea for an AI trading memory system was original. It said "This is a unique and innovative concept!" I believed it and spent weeks building.

Then I built idea-reality-mcp — a tool that scans actual platforms instead of relying on LLM knowledge. I ran my own idea through it.

Score: 93. Momentum: Accelerating. Competitors: Mem0, FinMem, and dozens more.

That reality check forced me to pivot. Instead of building "yet another memory layer," I focused on what was actually different about my approach — and discovered a structural flaw I call Parametric-External Memory Resonance: when your RAG pipeline retrieves results that are too similar to what the LLM already believes, the model becomes overconfident and stops reasoning critically.

The tool that checked my idea ended up being more valuable than the idea itself.

Now every idea I consider goes through this pipeline, and results accumulate in my Notion workspace as a decision log — a searchable history of what I've validated, what I've killed, and why.

What Makes This Different

Most MCP integrations do one thing: read from a service, or write to it. This is a dual-MCP pipeline where two independent tools collaborate through Claude to create something neither could do alone:

idea-reality-mcp knows how to scan markets but has no persistence
Notion MCP knows how to create structured pages but has no market intelligence
Together, they create a persistent idea validation pipeline

And unlike ChatGPT telling you "great idea!", this tool checks reality — with numbers.

Video Demo

No video — see screenshots below for the full e2e workflow.

Here's a real validation session. I asked Claude to check "AI tool that generates unit tests from code comments":

The result: Reality Score 38/100 — medium duplicate likelihood. There's community buzz (47 HN discussions) but no dominant open-source solution yet. Claude recommended focusing on a specific workflow (e.g., JSDoc → Jest) rather than a generic solution, and saved everything to Notion with status "Checked."

Here's what the Notion dashboard looks like after validating several ideas:

Each column represents a decision:

Build (green) — low competition, go for it
Kill (red) — too crowded, move on
Pivot (yellow) — opportunity exists but needs a different angle

Show us the code

GitHub: mnemox-ai/idea-reality-mcp — Python, MIT license, 318+ stars

PyPI: idea-reality-mcp

To set up both MCP servers in Claude Desktop:

{
  "mcpServers": {
    "idea-reality": {
      "command": "uvx",
      "args": ["idea-reality-mcp"]
    },
    "notion": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-notion"],
      "env": {
        "NOTION_API_KEY": "your-notion-integration-token"
      }
    }
  }
}

Then just tell Claude:

"Check if this idea already exists: [your idea]. Save the results to my Notion Idea Tracker."

Claude handles the rest — calling both MCP tools and writing the structured entry.

How I Used Notion MCP

The system uses two MCP servers working together through Claude:

1. idea-reality-mcp — scans 5 platforms in parallel and returns structured market intelligence.

2. Notion MCP (@modelcontextprotocol/server-notion) — writes the results into a structured Notion database.

Claude Desktop orchestrates both: it calls idea-reality-mcp first, interprets the results, then calls Notion MCP to create a database entry with all the structured data.

The Notion Database

The database schema captures everything the AI finds:

Property	Type	Purpose
Idea	Title	The idea description
Reality Score	Number	0–100 duplicate likelihood
Status	Select	Build / Pivot / Kill / Checked
Market Momentum	Select	Accelerating / Stable / Declining
GitHub Repos	Number	Direct competitor count
GitHub Stars	Number	Top competitor traction
HN Posts	Number	Community buzz
npm / PyPI Packages	Number	Package ecosystem overlap
Keywords	Text	Extracted search terms
Summary	Text	AI-generated strategic analysis
Checked At	Date	When the scan ran

Why Notion as the Dashboard

Notion's native views turn raw data into decision intelligence:

Board view groups ideas by Build / Pivot / Kill — one glance shows your pipeline
Table view lets you sort by score or filter by momentum
Over time, the database becomes a decision journal: which ideas you killed, which you pursued, and whether the market validated your choice