FutureSearch

Name: FutureSearch
Author: Unclaimed listing

AI & agents

AI forecasting with a public track record.

Unclaimed listing

Found via hacker-news ↗

Visit ↗

Unclaimed listingDid you build FutureSearch?

Verify ownership to edit the listing, publish updates, access analytics, and use launch tools.

What it does

FutureSearch is an AI forecasting product that helps you answer forward-looking questions with probabilistic, numeric, date, categorical, conditional, and decision forecasts. It’s built for teams and individuals who need defensible predictions—whether you’re evaluating market outcomes, company performance, policy/geopolitical scenarios, or internal business decisions—and want a forecasting workflow supported by research and repeatable methods.

You start by entering a question and selecting the forecast type (e.g., a probability like “Will 2026 be the hottest year on record?”, a numeric estimate like “What will Anthropic’s ARR be at year end?”, a date like “When will OpenAI IPO?”, or a conditional/decision question tying outcomes to a scenario). FutureSearch then runs the forecast and provides supporting research as part of the workflow, with tooling designed around producing a forecast—described as including multi-agent and research team capabilities.

FutureSearch emphasizes a public performance record and evaluation results, including a live tournament standing on Metaculus FutureEval and a benchmark placement on ForecastBench. It also publishes real-money trading results and positions on markets such as Kalshi and Polymarket, alongside S&P 500-related results, with research behind positions published (including losing positions).

For developers and power users, FutureSearch offers an API plus SDK/docs and integrations via an MCP endpoint that can be connected as a custom connector in Claude.ai / Claude Desktop, and used from tools like Claude Code (via a CLI command). Practical outcomes include turning ad hoc “what happens next?” questions into structured forecasts with attached research, and enabling forecasting as a callable capability inside agentic or developer workflows via API/MCP integration.

Pricing is usage-based per question, listed as $0.15 to $2 per question depending on effort, and there is a “Try it for free” option noted on the product along with a research plan promotion (“half price for your first month”).

Launch details