insum
insum is an AI research platform that runs entirely on the user's own hardware, transforming complex investigations into structured reports with an interactive Evidence Atlas. The platform requires user approval of a research plan before execution and allows mid-execution steering, ensuring every claim is mapped to its original source and single-source conclusions are flagged.
- Data & Analytics
- AI Agents
- Artificial Intelligence
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Traditional search engines return lists of links, requiring users to sift through pages and manually verify claims. This process is time-consuming, leaves room for unverified or conflicting information, and raises privacy concerns when queries are processed on external servers.
Solution
Insum is a research engine that runs entirely on the user's own hardware, delivering structured research reports instead of link lists. The platform begins with preliminary reconnaissance and a research plan that the user must approve before any investigation occurs. During execution, insum cross-references all gathered evidence, maps each claim to its original source, and flags single-source conclusions so no finding is stronger than its evidence. Every investigation produces a written report with an executive summary and thematic analysis, plus an interactive Evidence Atlas to explore supporting quotations, conflicting findings, and research gaps. The engine discloses what was found, what could not be reached, and which findings are in tension with one another.
Target Audience
Primary users are researchers, analysts, and professionals who require verifiable, private research outputs without compromising on data sovereignty or accuracy.
Features
- Runs all services locally on user-owned hardware, eliminating external server dependencies
- Answer-first structured reports with executive summary, thematic analysis, and claim-to-source ledger
- Interactive Evidence Atlas showing supporting quotations, conflicting findings, source concentration, and research coverage
- Preliminary reconnaissance and investigation plan requiring user approval before research begins
- Mid-execution steering capability that allows direction changes without restarting
- Automatic flagging of single-source conclusions and explicit disclosure of inaccessible sources and contradictory data