XPOZ

**Product and the concrete problem it solves.** XPOZ addresses the information gap between public social activity and the organizations that need to make decisions from it. Social platforms contain early signals of brand crises, coordinated influence campaigns, fraud, political manipulation, cyber threats, and fast-moving public sentiment, but the data is fragmented across APIs, inconsistent schemas, rate limits, and manual browser workflows. XPOZ's current product is a developer-facing social data API, MCP server, SDK, and CLI that let software teams and AI agents query structured data from Twitter/X, Instagram, TikTok, and Reddit. The company presents this as a social data layer for AI: developers can build monitoring dashboards, crisis-alert bots, research tools, competitive-intelligence workflows, or threat-detection systems without separately engineering scraping and normalization pipelines for each network. Its earlier public positioning emphasized detection and mitigation of disinformation, coordinated digital attacks, and harmful content; the current site broadens that wedge into agent-ready social intelligence. The strategic problem is the same in both descriptions: turn noisy, adversarial, rapidly changing public conversation into usable evidence before a narrative or operational incident becomes obvious.

**Core technology and how it works.** XPOZ's architecture is a data and access layer rather than a single classifier. The official site describes billions of indexed posts, user profiles, engagement metrics, historical access, advanced filters, pagination, smart caching, and real-time queries. Its MCP server exposes those capabilities to compatible AI tools through streamable HTTP, while the TypeScript SDK wraps authentication, polling, pagination, and data operations for application developers. The system can therefore support both deterministic queries and agentic workflows: a customer might search a topic across platforms, identify influential accounts, compare engagement over time, map conversation networks, and route the output into an alert or investigation workflow. Public materials mention OAuth 2.1, field selection, event resumption, and token-based access, but do not disclose the underlying collection agreements, crawler architecture, model providers, ranking algorithms, entity-resolution methods, or precision and recall benchmarks. XPOZ's defensibility consequently depends on data coverage, freshness, lawful and durable access, normalization quality, and the usefulness of its network and narrative representations. The company should not be credited with proprietary foundation-model IP that it has not publicly described.

**Market, customers, and go-to-market.** XPOZ sells infrastructure to organizations that need social evidence but do not want to build a multi-platform data pipeline. The commercial use cases on its website span brand monitoring, social listening, marketing intelligence, lead generation, research, competitor analysis, crisis response, developer relations, legal and compliance work, and security investigations. Its MCP-first distribution is notable: an engineering team can connect an existing AI assistant or internal agent to XPOZ through a remote server, then expand usage through SDKs and a CLI rather than adopting a large standalone dashboard. That approach can reach developers, security teams, communications groups, researchers, and intelligence analysts through bottom-up experimentation. The company is based in Herzliya and Startup Nation Finder classifies it as a B2B startup in Media & Entertainment Technologies, while LinkedIn describes an 11–50-person organization. Public sources do not identify named enterprise customers, annual recurring revenue, retention, or geographic revenue mix. The go-to-market thesis is therefore plausible and broad, but still dependent on converting free or developer-led usage into contracts with durable data, privacy, and service-level requirements.

**Traction, funding, and third-party validation.** Startup Nation Finder reports that XPOZ was founded in June 2023 by Ran Farhi, Danielle Israel, Ori Shaashua, and Mohar Bar and raised $3.2 million in one pre-seed round. The same profile describes a platform using machine learning to detect and analyze coordinated campaigns and to support content-removal and reputational-protection workflows. XPOZ's own website now shows a functioning product surface with public documentation, examples, an MCP endpoint, SDK guidance, and a free sign-up path; the TypeScript documentation is dated April 2026 and identifies the product's multi-platform data coverage. LinkedIn lists XPOZ as founded in 2024 and describes an 11–50-person team, so the founding-year discrepancy should remain explicit rather than silently resolved. A 2025 Israel Internet Association stakeholder analysis places XPOZ among Israeli companies working on disinformation and inauthentic-activity detection, providing ecosystem-level corroboration of the earlier mission. These signals show a live product and credible early financing, not proof of large-scale commercial adoption. No public source reviewed establishes audited data volume, customer count, revenue, model accuracy, or a government contract.

**Founders and team background.** The public record names Ran Farhi as co-founder and CEO, Danielle Israel as co-founder, Ori Shaashua as co-founder, and Mohar Bar as co-founder and COO. XPOZ's public posts also reference Prof. Irad Ben-Gal as a founder or technical contributor discussing inauthentic campaigns, suggesting that the founding narrative may have evolved as the product and team were formalized; this should be clarified in diligence. The company has presented at Wikimedia Israel and Reichman University discussions on fake content and has published examples of coordinated narrative analysis, indicating subject-matter engagement beyond generic social listening. LinkedIn lists additional engineering staff but does not provide a complete leadership biography, security organization, or headcount breakdown. The team's evident strengths are the combination of Israeli data-science and influence-analysis expertise with an ability to package the capability as developer infrastructure. Missing evidence includes prior exits, experience operating high-volume data services, platform-partner relationships, privacy counsel, and the operational controls required when customers use social data for consequential decisions.

**Competitive dynamics.** XPOZ competes with several categories. Cyabra is an Israeli influence-intelligence company focused on inauthentic behavior, bot networks, and narrative analysis for governments and enterprises. ActiveFence and Spectrum Labs provide trust-and-safety, content-moderation, and harmful-behavior detection platforms. Brandwatch and Talkwalker offer mature enterprise social-listening suites with broad historical datasets and established marketing distribution. Dataminr and Graphika compete in real-time event and network intelligence, while direct API and scraping approaches remain an incumbent alternative for sophisticated engineering teams. XPOZ's stated edge is the combination of a developer-first API, MCP-native agent access, cross-platform historical and real-time coverage, and security-oriented recipes such as zero-day discussion monitoring and coordinated-campaign analysis. That positioning could make XPOZ easier to embed in new AI workflows than a dashboard-first incumbent. The edge remains conditional: data licensing or platform-policy changes can remove coverage, larger providers can add MCP connectors, and specialized influence firms may have deeper labeled datasets and government references.

**Defense, security, and resilience relevance.** XPOZ has credible dual-use relevance because information integrity is a resilience and security function as well as a commercial analytics function. Civilian customers can use it to detect brand impersonation, coordinated fraud, crisis narratives, and emerging reputational threats. Security and public-sector users could apply the same data layer to identify coordinated inauthentic behavior, foreign influence patterns, hostile narratives, cyber-threat discussion, or disinformation targeting an election, military organization, emergency service, or critical-infrastructure operator. The Israel Internet Association's disinformation-resilience work and XPOZ's own influence-defense positioning support this adjacency. The calibration matters: public sources do not establish a defense or intelligence customer, classified deployment, government accreditation, or a system for taking action against accounts. Social data can also be incomplete, biased, privacy-sensitive, and manipulated. XPOZ is best understood as enabling information resilience and investigative capacity, with a technically credible path to security use, rather than as a fielded defense system.

**Growth stage, trajectory, and key diligence risks.** XPOZ is early-stage: it has a live developer product and reported pre-seed capital, but no public evidence of scaled enterprise revenue or repeatable government procurement. Its trajectory depends on turning a difficult data-collection problem into a trusted substrate for AI agents and security workflows. Key diligence points are (1) platform dependence and terms-of-service or API-access changes; (2) privacy, consent, retention, and cross-border data-governance obligations; (3) false positives and analytical overreach when network signals are used to label people or campaigns; (4) the cost and reliability of indexing billions of posts while maintaining freshness; (5) competitive pressure from social platforms, enterprise suites, and specialized intelligence vendors; (6) abuse risk if the API is used for harassment, surveillance, or manipulation; and (7) a public-record gap around customers, revenue, data rights, model evaluation, and the exact founder/team structure. The most valuable next milestones would be disclosed enterprise references, independent validation of campaign-detection quality, durable platform coverage, audited security and privacy controls, and evidence that XPOZ's MCP and API distribution creates retained usage rather than short-lived experimentation.