Connect
Point your AI assistant at mcp.squark.ai and sign in with your Squark account. One connection; enterprise sign-in; your credentials stay yours.
The Squark MCP Server — A fundraising industry first
A Fundraising Industry FirstYour AI assistant just learned to predict…
Squark AI's predictive AI is now available inside Claude, ChatGPT, Gemini, and any MCP-compatible assistant. Ask in plain English. Get real predictions from your donor data. Act in the same conversation.
Works with Claude · ChatGPT · Gemini · any MCP client
Generative asks. Predictive answers.
Your assistant is brilliant with language — and guessing about your donors. Squark is the inverse: a decade of predictive AI for fundraising, no prose required. MCP is the handshake. The question in your words; the answer from your data.
How it works
Three steps — connect, ask, act.
Ask
In plain English: "Score our lapsed donors for reactivation likelihood." Your assistant calls Squark's tools — register data, build the model, run the scoring.
Act
Ranked scores, explainable drivers, audiences, and forecasts return to the conversation — where the same assistant drafts the appeal and cuts the segment.
Connect
Connect your assistant
Three steps, one URL — the same MCP endpoint in every client.
- Open Claude and go to [VERIFY exact menu path] → Connectors.
- Choose Add MCP server and paste the URL below.
- Sign in with your Squark account when prompted.
MCP endpoint
Enterprise sign-in with your Squark account. IT can review the connection in one call.
The signature story
From question to campaign…
Maria is not trying to "use AI." She is trying to get a campaign out the door.
- 9:02"Register last year's donor file and this year's lapsed list." Done.
- 9:06"Build a model predicting who gives again this fiscal year." Squark's AutoML trains while she pours coffee.
- 9:14"Score the lapsed list. Show the top two deciles — and why each donor ranks high." Ranked file, drivers included.
- 9:21"Draft a reactivation email for the top decile, in our voice." The generative side takes over.
- 9:30Audience, insight, and copy — before her first meeting.
A three-week analytics ticket, closed over one cup of coffee.
What you can ask
Real prompts, real predictions.
Score our lapsed donors for reactivation likelihood.
Build an upgrade model and show me the top decile.
Which sustainer prospects should we call this month — and why?
Score this acquisition list before we mail it.
Forecast Q4 giving from the current pipeline.
Explain why donor #48211 ranks so high.
Division of labor
Why frontier models don't solve prediction alone.
A frontier model can talk about donor likelihood.
Squark can calculate it.
A frontier model can summarize a file.
Squark can score it.
A frontier model can draft the appeal.
Squark can identify the audience.
The risk is not that frontier models are weak. The risk is that they are persuasive. The better architecture: the assistant orchestrates and explains; Squark predicts. That is exactly what MCP enables.
What this replaces
Three patterns this ends.
A
Dashboard chat
A chat window on reports is still descriptive — it tells you what the dashboard already knows. It cannot register a file, build a model, or score an audience.
If the AI can't run the model, it's reading you the weather — not forecasting it.
B
Shadow AI
Exporting donor CSVs into a generic chatbot feels powerful — but an LLM is not a governed prediction workflow, and donor PII in consumer tools is exactly what IT fears.
That's not predictive AI; that's a confident guess with a compliance problem.
C
Wait-and-see
Caution is reasonable. But the next appeal is going out either way — the only question is whether the list comes from a model or a guess.
Your next appeal is going out either way. The only question is who's on the list.
Use cases
Six ways teams put this to work.
Lapsed donor reactivation
Rank the file; focus budget on donors most likely to return.
Upgrade targeting
Find who's ready for a bigger ask — and why.
Sustainer growth
Spot monthly donors ready to grow, without over-soliciting.
Acquisition scoring
Score the list before the money is spent.
Revenue forecasting
Ground the Q4 number in data, not narrative.
Donor-level explanation
Every score comes with its why.
Trust
Built for trust.
Sign in with Squark
Enterprise authentication on AWS Cognito — your team's own accounts, never shared keys.
Isolated by organization
Per-user, per-org credentials; encrypted at rest; tokens rotate automatically.
Your data stays yours
Donor data never trains foundation models.
Human-in-the-loop by default
The assistant proposes; your team approves every consequential step.
The assistant proposesSquark computesYour team approves
We do not compete with the assistant.
We give the assistant something true to say.
Generative asks. Predictive answers.
The white paper
Generative asks. Predictive answers. — Why the Squark MCP Server changes fundraising AI
Sixteen pages on the category shift: why frontier models don't predict, what MCP changes, the Monday-morning workflow, six priority use cases, and the trust architecture — plus a one-page executive brief to share with your board.
Request a demo
See it on your data.
A 30-minute walkthrough of the Squark MCP Server inside your assistant — on a sample of your donor data.
FAQ
Questions we hear most.
Better fundraising starts with better predictions…
Bring one priority use case: reactivation, upgrade, sustainer growth, acquisition, or forecasting.