Device signals meet AI risk
Fingerprint’s MCP Server launched an AI-capable service that ingests device and browser events to flag live anomalies for fraud prevention — a tool increasingly embedded in payments, brokerages, and asset platforms. Combining device intelligence with portfolio-operational risk engines is emerging as a practical mitigation step against AI-enabled scams.
Fingerprint rolled out its Model Context Protocol (MCP) Server as an invitation-only preview with live demos scheduled at MRC Vegas, March 16–19, 2026, according to the company blog. (fingerprint.com) The product ships in two deployment modes—open-source software for on-prem teams and a managed service for cloud users—and both routes connect AI assistants to Fingerprint’s production APIs for operational workflows. (markets.financialcontent.com) Fingerprint’s Device Intelligence Report states the firm analyzed more than 23 billion device-identification events in its 2026 dataset. (fingerprint.com) Partners and integrations describe the platform surfacing over 100 browser/device/network signals per session for downstream decisioning. (businesswire.com) Adoption of the MCP concept has been rapid: downloads rose from roughly 100,000 in November 2024 to over 8 million by April 2025, reflecting broad developer interest in agent-to-tool plumbing. (helpnetsecurity.com) Commercial integrations show how device signals enter decision engines today—Unit21 announced a Fingerprint integration that feeds 100+ signals into its real-time rules engine to trigger blocks, challenges, or alerts, while Oscilar documented a Fingerprint partnership for banks and credit unions. (businesswire.com) AI-native risk vendors like Sardine position combined fraud/credit/AML decisioning stacks that can ingest MCP-delivered device intelligence for automated portfolio-operational controls and remediation workflows. (sardine.ai) Community activity includes public GitHub MCP Server implementations and an MCP registry listing connectors, indicating developer-driven interoperability options beyond single-vendor lock-in. (github.com)
