fans.vote MCP

Predict how people answer, from your AI assistant

https://fans.vote/mcp

claude mcp add --transport http fans-vote https://fans.vote/mcp --header "Authorization: Bearer <your key>"

Your key

Loading…

The tool: predict_poll_answers

Takes: question (3–500 characters), answers (2–20, all different), optional audience: brand (a Facebook page name), interest (page names or topics), country (ISO-2 codes), age_group (under 25, 25-34, 35-49, 50+), gender (male or female). interest, country and age_group take one value or a list.
Returns: answers[] (answer, percent, rank), winner, too_close, confidence.score (0–100) and confidence.components, baseline (no audience data, with shift), audience_used (per field: specific, fallback or none, and its share), evidence (people, similar polls, top_similar with each poll's public result), and exact_match: the real result when the same question and answers were asked before.
Too narrow: An audience that cannot stand on 100 similar polls returns “Not enough evidence”.
Limits: 20 calls a minute per connection address; about 300 a day per account.

Try these

Pick a slogan

Which tagline your audience prefers.

Use fans.vote predict_poll_answers to test these slogans for <brand> with <audience>: <slogan 1>, <slogan 2>, <slogan 3>. Report the winner, each share, the confidence, and which audience data was used or fell back.

Name a product

Rank candidate names before launch.

With fans.vote predict_poll_answers, ask "Which name fits <product> best?" with the answers <name 1>, <name 2>, <name 3>, for <audience>. Rank the names and tell me how confident the prediction is.

Choose a campaign angle

See which message lands.

Use fans.vote predict_poll_answers to compare these campaign angles for <product> with <audience>: <angle 1>, <angle 2>, <angle 3>. Show the split, the baseline shift and the closest real polls behind it.

Message per market

Compare one choice across countries.

Run fans.vote predict_poll_answers on "<question>" with the answers <answer 1>, <answer 2>, once for each country: <country codes>. Put the winners and confidence side by side.

Split by audience

Men vs women, or age groups.

Use fans.vote predict_poll_answers on "<question>" with <answers>, once for women and once for men (then once per age group). Tell me where they disagree and how confident each result is.