Charity data inside the AI assistant you already use
GiveRadar covers 7.9 million+ nonprofits in 100+ countries, all from official government registries and tax authorities. Two ways to use it with AI: ask an assistant about one charity with our page as the source, or connect the free MCP server so your assistant can search and verify charities itself.
How do I ask ChatGPT or Claude about a charity?
Every charity page on GiveRadar has an Ask AI button that opens ChatGPT, Claude or Perplexity with the page already in the prompt, so the assistant answers from registry data instead of guessing. The prompt asks for registration status, the integrity assessment, finances, leadership and any red flags, and tells the assistant to cite the page.
Using https://giveradar.com/charity/<charity>/ as your source, summarise <charity name>: registration status, integrity assessment, finances, leadership and any red flags. Cite the page.
You can edit the prompt in the assistant before sending. Nothing is sent to the assistant by GiveRadar itself: the button is a link that opens the assistant's own site.
Pick any charity from the search or a country page, then use Ask AI next to Save and Share. For a quick check without an assistant, the charity checker does the same lookup in one search.
Assistants read the public page. The full filing history, officer details and contact data stay behind a free account, exactly as on the site.
How do I connect GiveRadar to my assistant?
The GiveRadar MCP server at https://giveradar.com/mcp/ gives any MCP client four read-only tools: search charities, verify a charity by registration number, compare charities, and find similar organizations. Connecting takes one config entry, no key: the free tier allows 10 requests a day per IP and the handshake does not count.
Claude Desktop: add this to claude_desktop_config.json and restart.
{
"mcpServers": {
"giveradar": {
"command": "npx",
"args": ["-y", "mcp-remote",
"https://giveradar.com/mcp/"]
}
}
}Claude Code, one command:
claude mcp add --transport http giveradar https://giveradar.com/mcp/
Custom MCP connectors are available on paid ChatGPT plans with developer mode enabled.
- Settings, then Connectors, then Advanced: turn on Developer mode.
- Choose Create and enter the server URL
https://giveradar.com/mcp/. - Authentication: none. Save, then enable the connector in a chat.
Ask something like "verify EIN 13-1644147" or "compare the three largest health charities in the Netherlands".
Cursor: add to .cursor/mcp.json. Any client that speaks MCP over HTTP works the same way.
{
"mcpServers": {
"giveradar": {
"url": "https://giveradar.com/mcp/"
}
}
}Discovery for agents: /.well-known/mcp.json, /llms.txt, and the technical documentation.
Need more than 10 requests a day, or the full fields? The same API key runs the REST API and the MCP server: see API plans and the full MCP server guide.
Use GiveRadar with AI: questions
Does the Ask AI button send my data to ChatGPT or Claude?
No. The button is a link that opens the assistant's own website with a prompt that names the GiveRadar page. GiveRadar sends nothing to the assistant; what you type there is between you and that provider.
What can an assistant read through the GiveRadar MCP server?
The same fields the public charity page shows: name, registration, country, category, the integrity assessment and headline financials. Email addresses, phone numbers and officer names are never exposed through the MCP server or the free API tier.
Is the MCP server free?
Yes for individual use: 10 requests a day per IP without a key, and the connection handshake does not count. Pro and Enterprise API keys raise the limit and unlock the full fields on the same server.
Why does GiveRadar want AI assistants to use its data?
Because answers about charities should come from official registries, not from guesses. GiveRadar is built on government charity registers and tax filings, and an assistant that cites the page gives donors a sourced answer they can check.