Let Claude see what Google's Knowledge Graph holds for any brand, person or organization, how that entity connects to the real world, and what it still needs to earn a stronger knowledge panel.
Works with Claude Desktop, Claude Code, Cursor, VS Code and any other MCP client. Free and MIT-licensed. Uses a free Google API key, or Wikidata only without one.
Overview
The server gives your AI assistant four read-only tools. Each one pulls live data from the Google Knowledge Graph Search API and Wikidata.
find_entities
Lists every Knowledge Graph entity matched for a name, with its ID (e.g. /m/0dr90d), type, description and Google's result score, highest first.
analyze_entity_prominence
Ranks same-name entities on a transparent 0–100 prominence score and explains each score in plain English.
compare_entities
Compares 2–5 brands or people side by side: prominence, readiness, what each is missing, key connections and official profiles, and who leads on each signal.
get_entity
Returns a full profile: knowledge panel data, connections (founders, CEO, HQ, products, parent company…), official profiles, an 8-point readiness checklist and ready-to-paste JSON-LD.
Under the hood
Setup
You need Node.js 18 or newer. The first run downloads and builds the server, which takes about a minute.
Without a key, the server uses Wikidata only. That covers well-known entities, but many people and smaller brands exist in Google's Knowledge Graph and not in Wikidata.
Replace your-key-here with your key. Your key stays on your computer: it's only sent to Google's API.
Claude Code: run this in your terminal
claude mcp add knowledge-panel -e GOOGLE_KG_API_KEY=your-key-here -- npx -y github:theseoriddler/google_knowledge_panel_mcpClaude Desktop and Cursor
In Claude Desktop, open Settings → Developer → Edit Config, add this block, then fully restart the app. In Cursor, add it to ~/.cursor/mcp.json.
{
"mcpServers": {
"knowledge-panel": {
"command": "npx",
"args": ["-y", "github:theseoriddler/google_knowledge_panel_mcp"],
"env": { "GOOGLE_KG_API_KEY": "your-key-here" }
}
}
}VS Code (GitHub Copilot)
Add this to .vscode/mcp.json.
{
"servers": {
"knowledge-panel": {
"type": "stdio",
"command": "npx",
"args": ["-y", "github:theseoriddler/google_knowledge_panel_mcp"],
"env": { "GOOGLE_KG_API_KEY": "your-key-here" }
}
}
}Ask: “Use find_entities to look up nike.” If the reply's source says Google Knowledge Graph + Wikidata, your key is working. If it says Wikidata only, check the key in your config and restart the app.
Try it
Scoring
The readiness score is the share of these 8 checks an entity passes. Without an API key, the three Google checks are skipped.
A transparent 0–100 heuristic, not a Google metric. Signals that aren't available are left out and the rest re-weighted.
| Signal | Weight | Full marks |
|---|---|---|
| Google result score, relative to the top match | 40 | Top match |
| Knowledge panel fields (description, detailed source, image, website) | 20 | All 4 |
| Wikipedia language editions | 20 | 100+ |
| Wikidata statements | 10 | 1,000+ |
| Linked official profiles | 10 | 5+ |
Output
get_entity builds Organization or Person JSON-LD with sameAs links from the entity's actual graph data. Here's the output for Tesla (/m/0dr90d), trimmed. Review the links before you publish.
{
"@context": "https://schema.org",
"@type": "Organization",
"@id": "https://www.tesla.com/#organization",
"name": "Tesla",
"description": "American automotive, energy storage and solar power company",
"url": "https://www.tesla.com/",
"logo": "https://commons.wikimedia.org/wiki/Special:FilePath/Tesla%20Motors.svg",
"sameAs": [
"https://en.wikipedia.org/wiki/Tesla,_Inc.",
"https://www.wikidata.org/wiki/Q478214",
"https://www.google.com/search?kgmid=%2Fm%2F0dr90d",
"https://x.com/Tesla",
"https://www.linkedin.com/company/tesla-motors/",
"https://github.com/teslamotors"
]
}Google's search doesn't always return the obvious entity. For example, salesforce returns Salesforce Marketing Cloud, not Salesforce, Inc. compare_entities works around this by preferring exact name matches. With the other tools, run find_entities first and pass the right ID.
Connections and official profiles come from Wikidata. If an entity has no Wikidata item, those sections are empty, and adding one is one of the most effective ways to strengthen an entity. This project isn't affiliated with or endorsed by Google. Full documentation and troubleshooting are in the GitHub README.