What is FutureSearch?
FutureSearch is an AI forecasting service that answers questions about what will happen next. Ask for a probability, a number, a date, or which option wins, and its research agents search the web, read sources, and return a structured answer. It runs in a web app, through an API and Python SDK, or inside Claude.
Top Features:
- Question types: forecast probabilities, numbers, dates, which option wins, or open-ended research.
- Decision forecasts: compare likely outcomes for each option, including doing nothing at all.
- Agent Map: add a researched column to every row of a list or dataset.
Use Cases:
- Strategy planning: estimate when a competitor or market will reach a key milestone.
- Market trading: find prediction market trades where forecasts differ from current prices.
- List enrichment: research hundreds of companies at once and add answers as columns.
Who Can Use FutureSearch?
- Analysts and investors: put numbers on uncertain outcomes before committing money or resources.
- Policy researchers: estimate the effects of proposed rules, such as downtown congestion charges.
- Developers: add forecasts to pipelines through the API, Python SDK, or MCP.
Pricing
- Free ($0): $20 in free credit, ten researchers at a time, and API access.
- Analyst Starter ($20 per month): 15 percent off usage rates and 20 researchers at once.
- Research ($99 per month): 30 percent off usage, 40 researchers at once, and priority support.
Pros and Cons
Pros:
- Track record: ranks near the top of ForecastBench and Metaculus forecasting benchmarks.
- Pay per question: costs follow the effort level you pick, from cents to dollars.
- AI tool support: connects to Claude, Cursor, Codex, and Gemini CLI through MCP.
Cons:
- Variable costs: token-based billing makes the cost of big jobs hard to predict.
- No certainty: forecasts are probabilities, so they still need human judgement before acting.
- Developer leaning: advanced pipelines need Python 3.12 or an AI coding assistant.
FAQs:
1) What does a forecast cost?
Low-effort forecasts cost about $0.25 to $3, and high-effort ones $2 to $5.
2) Is there a free trial?
Yes, new accounts get $20 in free credit with no card required.
3) Does it work inside Claude?
Yes, an MCP connector works with Claude.ai, Claude Desktop, and Claude Code.
4) What is multi-agent mode?
It sends one question to three to six agents, then merges their findings.
5) What do larger teams pay?
Expert costs $500 monthly for 80 researchers, and enterprise pricing is custom.