Hyperresearch

Product Introduction

1. Definition: Hyperresearch is an AI-powered research automation platform and agent that conducts deep, evidence-based research. It falls into the technical categories of AI research agents, automated due diligence tools, and citation-verification systems.
2. Core Value Proposition: It exists to solve the problem of AI hallucination and shallow research by automating the creation of fully cited, source-verified reports. Its core value is durable, verified AI search, where every claim is checked against its source and all evidence is stored in a persistent, searchable knowledge vault.

Main Features

1. Evidence Vault: Every source fetched during a research run is permanently stored in the user's private vault. This includes the full text (where legally available), provenance metadata, timestamps, and the exact quotes supporting each claim. The vault is semantically searchable, and subsequent research queries search the vault first, reducing cost and improving context. It supports export and integration via MCP (Model Context Protocol) and API.
2. Multi-Source Research Engine: The platform aggregates over 40 structured data APIs and 4 AI search engines. It accesses the open web (with logged-out and browser-rendered fetches), scholarly databases (OpenAlex, PubMed, Crossref), official records (SEC EDGAR, U.S. Census, BLS), and social platforms (X) for real-time commentary. For paywalled academic papers, it actively seeks legal open-access copies via Unpaywall, Europe PMC, and CORE to read the full text, not just abstracts.
3. Citation Verification & Receipt: This is the platform's signature feature. After drafting a report, the system automatically checks every cited claim or figure against the verbatim text of its source. It generates a "verification receipt" with per-citation verdicts (e.g., supported, partially-supported, unsupported), flags retracted papers, and corrects inaccuracies before the report is delivered. A separate Verification API can audit third-party documents.
4. Staged Research Process (Deep Run): The "Deep" research mode employs a rigorous, multi-stage process: decomposing the question, conducting wide sweeps, identifying contradictions ("fight clusters"), performing targeted depth investigations ("loci analysis"), planning the report based on an evidence map, writing in a single pass, and finally verifying all citations. This contrasts with the faster "Light" mode, which uses a streamlined sweep-draft-edit-verify pipeline.
5. Answers MCP & API for AI Agents: This feature allows other AI agents (e.g., in Claude, Cursor, ChatGPT) to offload research. Instead of an agent performing its own unreliable web search, it can call Hyperresearch's ask function via MCP to receive a concise, verified answer with supporting quotes, or a definitive "not found." The search function returns ranked source results. This prevents search loops and hallucinations within the agent's context window.

Problems Solved

1. Pain Point: AI Hallucination and Unverified Citations. Standard AI assistants and search tools often invent sources or misrepresent data. Hyperresearch directly solves this by mechanically verifying every citation against the source text before delivery.
2. Pain Point: Ephemeral and Non-Auditable Research. Typical web searches and AI chats provide answers without preserving the underlying evidence. Hyperresearch creates a durable, searchable evidence vault, turning one-off queries into a growing knowledge base.
3. Target Audience: Professionals requiring high-fidelity research. Key personas include: Due Diligence Analysts, Investment Researchers, Academic Researchers and Students, Investigative Journalists, Content Strategists needing verified data, and Developers building reliable AI agents that need factual grounding.
4. Use Cases:
   • Commercial Due Diligence: Researching a company's market position, financials (via SEC EDGAR), and competitive landscape with fully sourced reports.
   • Academic Literature Review: Quickly surveying a field, identifying key papers and scholarly debates, with automatic checks for retractions.
   • Agent-Augmented Development: Providing a reliable, citation-checking research backend for custom AI agents using the MCP or REST API.
   • Pre-Publication Fact-Checking: Using the Verification API to audit citations in an article, white paper, or report before publication.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):

   • Pros:
     • Unmatched Citation Integrity: The automated, pre-delivery claim check is a unique and critical strength for trust.
     • Evidence Durability: The vault transforms research from a disposable activity into a cumulative asset.
     • Source Breadth and Depth: Combines wide web search with deep, structured data APIs.
     • Agent-Native Design: The MCP server integration is a forward-looking architecture for the AI agent ecosystem.
     • Transparent Process: The detailed breakdown of the "Deep" research stages provides unusual visibility into its methodology.
   • Cons:
     • Cost and Speed: A "Deep" run costs $49 and takes 3-5 hours, making it unsuitable for quick, casual queries. It's a premium tool for high-stakes research.
     • Complexity: The wealth of features (vault, MCP, verification API) may have a steeper learning curve than simple search tools.
     • Dependency on Source Availability: Accuracy is contingent on its ability to fetch source texts; completely paywalled or non-digital sources remain a challenge.

2. Key Alternatives & Differentiation:

   • vs. Perplexity/You.com (AI Search Engines): These provide cited, real-time answers but are designed for speed and breadth. Hyperresearch differentiates through its systematic depth investigation, pre-shipment citation verification, and permanent evidence vault. Perplexity does not store your research history as a searchable knowledge base or verify claims post-drafting.
   • vs. Consensus or Elicit (Academic AI): These tools are specialized for querying academic literature. Hyperresearch is broader, combining scholarly search with web, financial, and government data into a single, verifiable report for professional use cases beyond pure academia.
   • vs. Manual Research + AI Writing (e.g., ChatGPT): A human-led process offers ultimate control but is extremely time-intensive. Hyperresearch automates the evidence gathering, synthesis, and verification at scale, providing a structured audit trail (the vault and receipt) that manual methods lack.

Frequently Asked Questions (FAQ)

1. How does Hyperresearch's citation verification work? The system extracts every claim that cites a source from the draft report. It then fetches the exact source text and compares the claim against it word-for-word. It assigns a verdict (supported, partially-supported, etc.), makes guarded corrections if needed, and produces a verification receipt. This happens automatically before any report is delivered to the user.
2. What is the difference between Hyperresearch's Light and Deep plans? The Light plan ($9/run) is a faster, streamlined process suitable for straightforward questions, taking ~30-40 minutes and producing a 2.5k-4k word report. The Deep plan ($49/run) is a comprehensive, multi-stage investigation for complex topics, taking 3-5 hours, consulting 150-200+ sources, and including the full verification receipt and deeper analytical stages like contradiction analysis.
3. Can I use Hyperresearch to check citations in my own existing documents? Yes, via the standalone Verification API. You can submit a document and its source list. Hyperresearch will analyze it, provide per-citation verdicts, check quote integrity, sweep for retracted papers, and flag duplicate sources, functioning as an automated pre-publication audit tool.
4. How does the "vault" feature benefit me over time? The vault acts as your personal, searchable research database. Every source from every run is stored indefinitely. When you ask a new question, the system searches your vault first, leveraging your past research to provide cheaper, faster, and more context-aware answers, effectively making your research process cumulative and more efficient.
5. What is Answers MCP and how do I connect it to my AI agent? Answers MCP is a remote MCP server that exposes your Hyperresearch vault and research capabilities to AI agent clients like Claude Desktop or Cursor. You authorize the connection via OAuth in your browser. Your agent can then use the ask tool to get verified, concise answers or the search tool to retrieve raw sources, preventing the agent from performing unreliable web searches itself.