Guide · Model Context Protocol
What is MCP, and why should fundraisers care…
Model Context Protocol — MCP — is the plumbing that lets Claude, ChatGPT, and Gemini work with your donor data safely. Here is what it is, in plain English, and why it changes what an AI assistant can do for a development team.
MCP in one sentence
Model Context Protocol (MCP) is an open standard that lets AI assistants — Claude, ChatGPT, Gemini — talk to your tools and data through a shared interface, instead of every vendor building a one-off integration.
What an MCP server actually is
An MCP server is a small backend that publishes a defined set of tools the assistant can call. Each tool has a contract: what it does, what it needs, what it returns. The assistant can only use the tools the server exposes — it cannot invent new ones or reach past them.
Why this matters for fundraising
Fundraisers already ask AI assistants to draft appeals, summarize board reports, and clean up copy. MCP is what lets the same assistant do the quantitative work too — likelihood-to-give, lapsed-donor reactivation, revenue forecasting — on your donor data, in the same conversation, without your team leaving the assistant they already use.
How donor data stays yours
Donor data goes to the MCP server, not to the foundation model. Your team signs in with their own accounts; credentials scope per user and per organization. The assistant proposes a step; a human on your team approves it. The answers returned to the chat are predictions and explanations — not raw records handed to a model provider to train on.
Where predictive AI fits in
A general-purpose assistant is good at language. It is not, on its own, a modeling engine — it will happily guess numbers if you let it. An MCP server built on a real predictive platform makes the assistant honest: every score comes from a model trained on your data, and every explanation traces back to the drivers behind that specific prediction.
What a fundraising MCP workflow looks like
Register a donor file. Pick the outcome to predict (major-gift likelihood, lapse risk, next-gift amount). Build the model. Score a fresh file. Retrieve ranked results. Ask the assistant to explain why donor #14 scored where they did. Every step is a tool call your team approves, not a black-box answer.
Common questions
MCP, answered
Is MCP a Squark thing or an open standard?
MCP is an open standard, originally introduced by Anthropic and adopted across AI assistants. Any vendor can publish an MCP server; any MCP-aware assistant can call one.
Do I need to run any of this myself?
No. The Squark MCP Server is hosted. Your team signs in with their Squark account inside their assistant of choice and starts asking.
Is this a chatbot on top of dashboards?
The opposite. The real engine — data registration, model building, scoring, explanations — is exposed as tools the assistant reads, writes, and executes. It never invents a number.
What can the assistant decide on its own?
Nothing consequential. Launch configuration is supervised: the assistant proposes, your team approves. An expanded autonomous tool set exists for teams that opt in.