FutureSearch
FutureSearch is an AI-native platform that orchestrates teams of LLM-powered forecasting and research agents across entire datasets, serving frontier AI labs, enterprises, developers, and analysts with calibrated predictions, multi-angle research, and structured data operations via API and MCP integrations.
- Company typePrivate
- Founded2023
- HeadquartersSan Bruno, United States
- Headcount1–10
- GTM typeB2B
- OfferingSoftware
What FutureSearch does
FutureSearch (legal entity Varuna AI, Inc., d/b/a FutureSearch) is a private AI research and forecasting company founded in 2023 in San Bruno, California, by Dan Schwarz and Lawrence Phillips, both formerly of Metaculus. The company provides a platform of orchestrated LLM-powered agents that operate on tabular data, enabling users to run forecasting, multi-angle research, parallel web research, ranking, classification, and entity-resolution operations across an entire DataFrame rather than one question at a time. Its technical differentiator is a proprietary multi-agent architecture combined with elite human forecasters and public accuracy benchmarks (BTF-2 and Deep Research Bench), under which FutureSearch was the first AI forecaster to beat humans in a Manifold Markets tournament (December 2023) and the first to correctly break down OpenAI's revenue when public estimates were inaccurate (July 2024).
FutureSearch monetizes through usage-based, per-question pricing: a batched forecaster at $0.15-$0.50 per question and a high-effort forecaster at $1.60 per question, with a $20 free credit to start. Prior to its April 2026 public launch, the company generated revenue primarily by selling bespoke forecasting and research intelligence to frontier AI labs, including Google, Anthropic, OpenAI, and xAI. The platform is now distributed self-serve via web app, Python SDK, and MCP server, with native integrations into Claude.ai, Claude Code, Cursor, ChatGPT Codex, Google Sheets, and Microsoft Excel. A companion sub-product, Stockfisher, provides AI-generated company revenue and margin forecasts.
The company is venture-backed, having raised a $5.7M seed round in November 2024 led by Metaplanet (with Juniper, Macroscopic, Ought, and TwinPath), and operates with a team of 12 including 7 PhDs. FutureSearch also publishes research artifacts (the AI 2027 timeline forecast, arXiv papers, and public benchmark leaderboards at evals.futuresearch.ai) to anchor category mindshare for AI forecasting and research-agent evaluation.
FutureSearch firmographics
Firmographics- Name
- FutureSearch
- Legal name
- Varuna AI, Inc.
- Website
- https://futuresearch.ai
- Company type
- Private
- Founded year
- 2023
- Operating status
- Operating
- Headcount range
- 1–10 employees
- Short description
- FutureSearch is an AI-native platform that orchestrates teams of LLM-powered forecasting and research agents across entire datasets, serving frontier AI labs, enterprises, developers, and analysts with calibrated predictions, multi-angle research, and structured data operations via API and MCP integrations.
- Ownership category
- akta.pro rank
FutureSearch industry classification
Industry- Product category
- AI Research and Forecasting Platform
- NAICS
- Computer Systems Design and Related Services (54151), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518), Custom Computer Programming Services (541511)
- SIC
- Services-Prepackaged Software (7372), Services-Computer Programming, Data Processing, Etc. (7370), Services-Computer Programming Services (7371)
- akta.pro primary industry
- Scenario Planning, Forecasting Platforms & Planning Intelligence (HDAAAHAL)
- akta.pro secondary industries
- MLOps/LLMOps & Model Lifecycle Management Services (BPAEAHAH), Data Engineering, Integration & ETL/ELT Services (BPAEAHAB)
Keywords
Where FutureSearch is headquartered
LocationHeadquarters
- HQ city
- San Bruno
- HQ country
- United States
- HQ region
- North America
Markets served
FutureSearch business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Infrastructure, Marketing or Sales, Operations
Revenue model
- Forecasting and Research API: Revenue is generated through per-question pricing for AI forecaster and researcher services. The batched forecaster costs $0.15–0.50 per question while the high-effort forecaster costs $1.60 per question. This is a usage-based model where customers pay based on the number of questions or research tasks they submit.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Pay-as-you-go | Free Plan - $20 free credit to start |
| Usage-based | Pay-as-you-go | Batched Forecaster |
| Usage-based | Pay-as-you-go | High-Effort Forecaster |
Go-to-market motion1 record
Distribution channels3 records
Marketing channels7 records
FutureSearch product offering
Product offeringCore offering
FutureSearch operates an AI-powered forecasting and research platform that runs teams of LLM-based agents on every row of a dataset. Its core operations produce calibrated forecasts (binary, numeric, date, categorical), multi-agent research synthesis, parallel row-level web research, dataset ranking by natural-language criteria, classification, and semantic entity matching/merging. Customers access these capabilities via a self-serve web app, a Python SDK/REST API, an MCP server, and native integrations with Claude, Cursor, ChatGPT Codex, Google Sheets, and Microsoft Excel.
Product overview
FutureSearch is a unified AI research platform offering multiple specialized AI agent operations. The core platform provides six interconnected products: Forecasters (calibrated probability and numeric forecasting), Multi-Agents (coordinated team research), Agents (parallel row-level web research), Scorers (ranking datasets by external criteria), Classifiers (categorization by custom rules), and Matchers (semantic table joining). All operations take DataFrames as input and return structured results, enabling batch processing of research tasks at scale. The platform is accessible via a web app, a Python SDK, and an MCP server for AI assistant integration. A companion product, Stockfisher, provides specialized revenue and margin forecasting for companies. The company also publishes public research benchmarks (Bench to the Future for forecasters, Deep Research Bench for research agents) to track industry accuracy.
Differentiator
Problem solved
Functional benefit
Products and services
- FutureSearch Platform An AI-powered research platform that provides teams of forecasters and researchers for analyzing datasets, predicting outcomes, and conducting large-scale research. Users spin up specialized AI agents (Forecasters, Multi-Agents, Agents, Scorers, Classifiers, Matchers) to perform calibrated forecasting, multi-agent research, parallel row-level research, ranking, classification, and semantic merging on tabular data.
- Forecasters A specialized AI operation that produces calibrated forecasts and rationales for binary (probability), numeric (percentile distributions), date (percentile dates), categorical, and thresholded questions, powered by elite human forecasters enhanced with proprietary AI systems. Accuracy is measured on the public BTF-2 leaderboard.
- Multi-Agents A team-of-agents research system that answers one question from multiple research angles and synthesizes findings into a single structured result, with configurable effort levels (low, medium, high) or custom agent directions.
- Agents A web research agent operation that processes every row of a DataFrame in parallel, fetching external information and returning structured results for each entry to populate new columns.
- Scorers An AI operation that ranks and sorts an entire dataset by a natural-language scoring criterion; agents derive the scoring metric at runtime by searching the web and reasoning over external information, then return the DataFrame sorted accordingly.
- Classifiers An AI operation that categorizes every entity in a dataset by any specified criteria using scalable web research, supporting binary, multi-category, and pass/fail screening classifications with optional reasoning output.
- Matchers An LLM-powered merge and dedupe operation that joins two DataFrames and groups duplicate rows using semantic entity resolution instead of exact or fuzzy key matches, handling subsidiaries, regional names, abbreviations, and product-to-parent-company mappings.
- Stockfisher An AI-generated revenue and margin forecasting application for companies, built on a FutureSearch forecaster that collects and analyzes publicly available information from the web.
- FutureSearch Python SDK and REST API A Python SDK and REST API enabling programmatic access to all FutureSearch operations (forecast, multi_agent, agent_map, rank, classify, merge, dedupe) via DataFrames; includes an MCP server for AI assistant integration. SDK is open-sourced at github.com/futuresearch/futuresearch-python.
Quantifiable outcome
- First to use an AI forecaster to beat humans in a prediction market tournament (December 2023)
- +3 more outcomes
Companies that use FutureSearch
Customer profileNamed customers9 records
Segments4 records
Ideal customer profiles4 records
FutureSearch technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration7 records
AI capability7 records
Feature7 records
FutureSearch partnerships and signals
Strategic signalPartnerships
Five partnerships are on record, tiered core and minor.
- AnthropiccoreAI provider whose Claude models power FutureSearch's services. Anthropic's AI technologies are used to facilitate prompt features that users input through the service.
- OpenAIcoreAI provider whose models power FutureSearch's services alongside Anthropic, Google, and X.
- GooglecoreAI provider whose models (Gemini) power FutureSearch's services alongside Anthropic, OpenAI, and X.
- Claude (Anthropic)coreNative integration available for Claude.ai and Claude Code. Users can add FutureSearch as a custom connector or MCP server directly within Claude products.
- StripeminorThird-party payment service provider used for payment services including card acceptance and merchant settlement.
Scale indicators5 records
Recent moves6 records
Expansion highlights5 records
FutureSearch competitors and assessment
Company assessmentBroad incumbents
- Glean: Glean is an enterprise AI search and knowledge-management platform indexing internal company data. Overlaps with FutureSearch's enterprise research and dedupe use cases at much larger scale across internal (rather than external) data sources.
- AlphaSense: AlphaSense is an enterprise market intelligence platform that aggregates research reports, transcripts, and filings for institutional investors. A scaled incumbent in a market FutureSearch targets with forecasting and research agents, though AlphaSense uses traditional search rather than AI agents.
- Perplexity: Perplexity provides AI-powered answer engines and research assistants for consumer and enterprise users. Overlaps with FutureSearch's general research/forecasting capability but at much broader scale and a different go-to-market (consumer-first).
Direct peers
- Metaculus: Metaculus is a public forecasting platform that aggregates human predictions, and is the prior employer of both FutureSearch co-founders. Directly comparable on forecasting methodology and tournament formats, though FutureSearch applies AI to the same problem space.
- Exa: Exa (formerly Metaphor) is a neural search engine optimized for AI applications, offering embeddings-based web search used by agents and LLM workflows. Comparable to FutureSearch's agent_map and multi_agent operations in serving AI-native research.
- Hebbia: Hebbia applies multi-agent AI to financial and legal document analysis, producing structured outputs from large unstructured datasets. FutureSearch targets similar enterprise workflows (lead scoring, market research) with comparable agent-orchestration architecture.
- Tavily: Tavily is an AI research API purpose-built for LLM agents to search and retrieve web content. Like FutureSearch, Tavily exposes multi-step research as an API that developers and agent builders consume programmatically.
- Samaya AI: Samaya AI builds AI research agents for financial analysts and institutional investors, focusing on factual research over the web. Direct overlap with FutureSearch's multi-agent research and Stockfisher forecasting product.
Emerging players
- Surge AI: Surge AI is an AI data labeling and reinforcement-learning-from-human-feedback platform. Adjacent to FutureSearch's classify and dedupe operations, though Surge is more focused on training-data creation than research/forecasting.
- Snorkel AI: Snorkel AI provides programmatic data labeling and AI data development workflows for enterprises. Overlaps with FutureSearch's classify, rank, and dedupe operations on enterprise data preparation use cases.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights7 records
Customer concentration
FutureSearch social profiles
Digital presenceFutureSearch compliance and trust
Trust signalCompliance2 records
FutureSearch financial estimates
Financial estimateRevenue estimate
Valuation estimate
FutureSearch leadership team
Management profileNumber of profiles
Profiles3 records
FutureSearch funding detail
Funding detailFunding overview
Funding rounds1 record
Investors
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
FutureSearch M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about FutureSearch
What does FutureSearch do?
FutureSearch operates an AI-powered forecasting and research platform that runs teams of LLM-based agents on every row of a dataset. Its core operations produce calibrated forecasts (binary, numeric, date, categorical), multi-agent research synthesis, parallel row-level web research, dataset ranking by natural-language criteria, classification, and semantic entity matching/merging. Customers access these capabilities via a self-serve web app, a Python SDK/REST API, an MCP server, and native integrations with Claude, Cursor, ChatGPT Codex, Google Sheets, and Microsoft Excel.
Is FutureSearch a public or private company?
FutureSearch is a private company. It is classified as venture growth investor backed and is currently operating.
When was FutureSearch founded?
FutureSearch was founded in 2023. It employs 1 to 10 people.
Where is FutureSearch based?
FutureSearch is headquartered in San Bruno, United States, in the North America region.
How does FutureSearch make money?
One revenue line is on record: forecasting and Research API.
Who are FutureSearch's main competitors?
Broad incumbents on record are Glean, AlphaSense and Perplexity. Direct peers are Metaculus, Exa, Hebbia, Tavily and Samaya AI. Emerging players are Surge AI and Snorkel AI.
Does FutureSearch have an API?
Yes. FutureSearch provides a Python SDK that enables developers to programmatically invoke AI-powered research agents. The API supports operations including forecast (calibrated forecasting), multi_agent (team research), agent_map (parallel row-level research), rank (scoring), classify (categorization), merge (fuzzy joining), and dedupe (entity resolution). Authentication uses API keys, accessed via the web app at https://futuresearch.ai/app/. An MCP server is available for integration with AI assistants. Developer documentation is at futuresearch.ai/docs/api.
What industry is FutureSearch in?
FutureSearch's product category is AI Research and Forecasting Platform. Its primary akta.pro industry code is HDAAAHAL, Scenario Planning, Forecasting Platforms & Planning Intelligence, with a secondary code of BPAEAHAH, MLOps/LLMOps & Model Lifecycle Management Services. Its NAICS code is 54151 and its SIC code is 7372.