Fatebook MCP Server

An MCP (Model Context Protocol) server for Fatebook.io - a prediction tracking tool that helps you improve your calibration.

URL: https://fatebookmcp.val.run

Setup

Get your API key from fatebook.io/api-setup
Add FATEBOOK_API_KEY to your MCP client's environment variables
Connect with auth header (the client handles this)

Tools

Tool	Description
`create_question`	Make a new prediction with optional initial forecast
`list_questions`	View predictions (filter by resolved/unresolved/search)
`get_question`	Get details about a specific prediction
`add_forecast`	Update your probability estimate (0-1)
`add_comment`	Add notes or reasoning
`resolve_question`	Mark as YES, NO, or AMBIGUOUS
`edit_question`	Update title, resolve date, or notes
`delete_question`	Remove a prediction

Usage

With MCP Chat

Add this server in settings:

Name: fatebook
URL: https://fatebookmcp.val.run
Auth env key: FATEBOOK_API_KEY

With OpenAI Responses API

const response = await openai.responses.create({
  model: "gpt-4o-mini",
  input: "Create a prediction: Will I finish my project by Friday? 70% confident, resolve by 2024-12-20",
  tools: [{
    type: "mcp",
    server_label: "fatebook",
    server_url: "https://fatebookmcp.val.run",
    headers: {
      Authorization: `Bearer ${FATEBOOK_API_KEY}`,
    },
    require_approval: "never",
  }],
});

Example Prompts

"Create a prediction: Will it rain tomorrow? 40% chance, resolve by tomorrow"
"Show my unresolved predictions"
"I'm now 80% confident on prediction [ID]"
"Resolve [ID] as YES"
"Add a comment to [ID]: New evidence suggests this is more likely"

What is Fatebook?

Fatebook is a tool for tracking predictions and improving your calibration. By recording your forecasts and comparing them to outcomes, you can identify biases and become better at estimating probabilities.

Key concepts:

Predictions: Yes/no questions with a probability and resolve date
Forecasts: Your probability estimates (can update over time)
Resolution: The actual outcome (YES, NO, or AMBIGUOUS)
Calibration: How well your probabilities match reality