Finic launches Kairos, an open-source fraud platform combining language-based checks with agent-managed workflows

Finic says Kairos combines traditional models such as XGBoost with plain-language checks, which it says can help teams make decisions without large labeled datasets. Its MCP server lets coding agents define workflows, train models on a company’s data, and select the best-performing variant; Kairos can be self-hosted with Docker or used as a hosted service.

For teams with limited labeled data or data-science capacity, Kairos’s language-based checks and agent-driven workflow could reduce the work needed to adjust fraud decisioning. The article does not report benchmark results or explain how teams should validate agent-selected models, so performance and oversight requirements remain open questions.